Adapting Market Design to High Shares of Variable Renewable Energy

IRENA HEADQUARTERS P.O. Box 236, Abu Dhabi United Arab Emirates

ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY RENEWABLE VARIABLE OF SHARES TO HIGH DESIGN MARKET ADAPTING ADAPTING MARKET DESIGN www.irena.org TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY recommendations 4. Conclusions and 4.Conclusions and distributed resources and distributed 3. Distribution networks 3.Distribution market design market 2.Wholesale transition 1. Power sector 1.Power Executive Executive Summary 2017 © IRENA 2017

Unless otherwise stated, material in this publication may be freely used, shared, copied, reproduced, printed and/or stored, provided that appropriate acknowledgement is given of IRENA as the source and copyright holder. Material in this publication that is attributed to third parties may be subject to separate terms of use and restrictions, and appropriate permissions from these third parties may need to be secured before any use of such material.

978-92-9260-025-9 (PDF)

Citation: IRENA (2017), Adapting market design to high shares of variable renewable energy. International Renewable Energy Agency, Abu Dhabi

ABOUT IRENA

The International Renewable Energy Agency (IRENA) is an intergovernmental organisation that supports countries in their transition to a sustainable energy future, and serves as the principal platform for international co-operation, a centre of excellence, and a repository of policy, technology, resource and financial knowledge on renewable energy. IRENA promotes the widespread adoption and sustainable use of all forms of renewable energy, including bioenergy, geothermal, hydropower, ocean, solar and wind energy, in the pursuit of sustainable development, energy access, energy security and low-carbon economic growth and prosperity.

ACKNOWLEDGEMENTS

This report has been directed by Salvatore Vinci and Rabia Ferroukhi (IRENA) and has been authored by: Ignacio Pérez-Arriaga, Tomás Gómez, Carlos Batlle, Pablo Rodilla, Rafael Cossent, Ignacio Herrero, Inés Usera, Paolo Mastropietro (Comillas Pontifical University), Salvatore Vinci (IRENA); with the contribution of: José Pablo Chaves, Binod Prasad Koirala (Comillas Pontifical University); Federico Boschi; Bishal Parajuli (IRENA).

The report benefited from input and review of experts: Pedro Linares (Comillas Pontifical University); Matthew Wittenstein (International Energy Agency); Marco Foresti (ENTSO-E); Christian Redl (Agora-Energiewende); Lori Bird, Jenny Heeter (NREL); Luiz Augusto Barroso (Empresa de Pesquisa Energética); Ruud Kempener (European Commission); Luca Lo Schiavo, Massimo Ricci (AEEGSI); Hannelore Rocchio; Henning Wuester, Sakari Oksanen, Paul Komor, Alvaro Lopez-Peña, Divyam Nagpal, Michael Taylor, Emanuele Taibi (IRENA).

For further information: [email protected]. This report is available for download from www.irena.org/publications DISCLAIMER

This publication and the material herein are provided “as is”. All reasonable precautions have been taken by IRENA to verify the reliability of the material in this publication. However, neither IRENA nor any of its officials, agents, data or other third-party content providers provides a warranty of any kind, either expressed or implied, and they accept no responsibility or liability for any consequence of use of the publication or material herein.

The information contained herein does not necessarily represent the views of the Members of IRENA. The mention of specific companies or certain projects or products does not imply that they are endorsed or recommended by IRENA in preference to others of a similar nature that are not mentioned. The designations employed and the presentation of material herein do not imply the expression of any opinion on the part of IRENA concerning the legal status of any region, country, territory, city or area or of its authorities, or concerning the delimitation of frontiers or boundaries. FOREWORD

he world has embarked upon a major energy transition which is bringing about profound changes in the ways electricity is produced, distributed, and Tconsumed. Renewable energy is at the centre stage of this transition. Fuelled by decreasing costs and improving technologies, renewable energy deployment has grown at an unprecedented pace. Each year, renewable power capacity additions set a new record, with over 160 GW added in 2016 alone. Still, the energy transition needs to happen faster for the world to meet key sustainable development and climate goals. The change will not happen merely with adjustments to the existing energy system, but through a paradigm shift across economies, societies and communities. Accelerating this transition requires a rethinking of electricity markets in many Adnan Z. Amin aspects, a key one being the adaptation of their design and operation to support Director-General higher shares of variable renewables –solar and wind energy – as well as distributed International Renewable power generation. Energy Agency In this context, this report presents the latest knowledge on the different options for the adaptation of liberalised electricity markets. Building on case studies from advanced markets, it identifies policy and regulatory measures needed to accommodate variable and decentralised renewables, while ensuring high standards of efficiency, reliability and environmental stewardship. The report highlights that regulations governing the electricity market must adapt to rapidly evolving needs and conditions in a timely manner. Enhanced flexibility for future power systems is an essential consideration. Electricity markets need to be re-designed to integrate all available resources, reward flexibility, and promote long- term investment. Modular, decentralised power generation requires new approaches to network regulation, advanced grid management methods and innovative metering technologies. Power sector regulation and renewable energy policy must work in tandem. They must reflect each country’s circumstances, ensuring reliable services at reasonable prices while sharing system costs and benefits fairly and equitably. “Adapting Market Design to High Shares of Variable Renewable Energy” offers recommendations for the entire power supply chain, from wholesale markets and system operations to distribution networks and end users. ADAPTING MARKET DESIGN I am confident that this report will prove useful, particularly as policy makers and TO HIGH SHARES OF regulators strive to transform the world’s power systems to achieve a sustainable VARIABLE RENEWABLE Aenergy Renewa future forbl all.e Energy Roadmap ENERGY ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

TABLE OF CONTENTS

EXECUTIVE SUMMARY...... 10

01 POWER SECTOR TRANSITION...... 18 1.1 The energy transition and its implications for the power sector...... 19

1.2 Power sector organisation ...... 22

1.3 Renewable energy impact on the power sector operation...... 24 1.3.1 System impacts of variable renewable energy...... 25 1.3.2 Market impacts of variable renewable energy...... 26

1.4 Conclusions...... 30

02 WHOLESALE MARKET DESIGN...... 32

2.1 Introduction...... 33 2.1.1 Electricity markets, products and time frames...... 33 2.1.2 Why rethink the market now...... 35

2.2 Short-term markets ...... 38 2.2.1 Time frames of markets, dispatches and prices...... 40 2.2.2 Locational granularity of prices and schedules...... 50 2.2.3 Bidding formats: fitting new necessities and resources...... 52 2.2.4 Pricing and clearing rules...... 58 2.2.5 Rethinking reserve requirements and procurement...... 63

2.3 Balancing markets ...... 67 2.3.1 Balancing responsibility and imbalance settlement ...... 68 2.3.2 Balancing products...... 70 2.3.3 Reserve markets and pricing of the products...... 75

2.4 Long-term support mechanisms...... 75 2.4.1 Generation adequacy mechanisms and RES integration...... 77 2.4.2 RES promotion mechanisms and wholesale market integration...... 83

2.5 Conclusions and recommendations...... 89

03 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES. . 90 3.1 I ntroduction...... 91

3.2 Advanced solutions for distribution network planning and operation...... 93 3.2.1 Proactive grid connection approach and long-term network planning...... 95 3.2.2 From reinforcement-based solutions to an active network operation...... 98 3.2.3 Paving the way through demonstration and pilot projects...... 100 3.2.4 Conclusions and recommendations...... 104

4 3.3 Advanced regulation to encourage innovation and efficient distributed generation integration...... 104 3.3.1 S hort-term measures to mitigate the impact of high levels of distributed generation on distribution revenues ...... 106 3.3.2 Guidelines for smarter distribution regulation...... 107 3.3.3 Conclusions and recommendations...... 111

3.4 Self-consumption: tariffs and metering...... 112 3.4.1. Net-metering policies...... 113 3.4.2 Self-consumption and the missing money problem...... 115 3.4.3 Short term solutions to tackle the missing money problem...... 117 3.4.4 Towards a new tariff design approach ...... 117 3.4.5 The importance of metering arrangements...... 121 3.4.6 Additional policy considerations...... 121 3.4.7 Conclusions and recommendations...... 122

3.5 Future role of distribution companies...... 123 3.5.1 Moving from network managers to market facilitators and system operators ...... 124 3.5.2 Advanced metering to enable demand response and competitive retail markets...... 128 3.5.3 Distributed storage and ownership models...... 131 3.5.4 Business models for the deployment of electric vehicle charging infrastructure ...... 135 3.5.5 Conclusions and recommendations...... 138

04 LESSONS LEARNT AND POLICY RECOMMENDATIONS...... 140 4.1 Wholesale market design...... 141 4.1.1 Adapting short-term market design...... 142 4.1.2 Improving long-term signals without ruining short-term ones...... 144

4.2 Distribution networks and distributed energy resources...... 145 4.2.1 Adapting regulation under high shares of distributed generation...... 145 4.2.2 Encouraging distribution companies to act as system operators...... 147 4.2.3 Enabling distribution companies to act as neutral market facilitators...... 148 4.2.4 Promoting sustainable development of self-consumption and demand response...... 149

REFERENCES...... 154

APPENDIX: EUPHEMIA, THE PAN-EUROPEAN MARKET CLEARING ALGORITHM ...... 164

5 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

LISTS OF FIGURES, TABLES AND BOXES

LIST OF FIGURES Figure 2.5 Example of schedule adjustments in intraday market...... 44

Figure 1.1 Levelised cost of electricity Figure 2.6 Time frame of European intraday markets. 45 for utility-scale power (ranges and averages), 2010 and 2016. . .20 Figure 2.7 Intraday trading arrangements across Europe...... 46 Figure 1.2 Energy installation costs and cycle life of battery storage technologies, Figure 2.8 The time frame of US real-time markets. . 48 2016 and 2030...... 21 Figure 2.9 Average German system imbalance Figure 1.3 Solar PV installed capacity within for every minute of the day during California Solar Incentive (CSI) program the year 2011...... 49 against the evolution of average cost of solar PV capacity...... 21 Figure 2.10 Approximation of the realized load and forecast errors on intraday markets. . 50 Figure 1.4 Traditional power sector organisation with a vertically integrated utility in each Figure 2.11 Simulation of suitable zones for zonal territorial franchise...... 23 pricing in Europe (maximum wind). . . . .51 Annual total of physical cross-border flows Figure 1.5 Competition wholesale market Figure 2.12 and regulated distribution companies in the CWE region and at the German acting also as retailers ...... 24 borders in 2016 (in TWh)...... 53 Monthly volumes of netted imbalances Figure 1.6 Fully liberalized power sector Figure 2.13 with competition at wholesale for all IGCC members for the imbalance and retail level with bilateral contracts. . .24 netting process...... 54 Evolution of the activation of the minimum Figure 1.7 Weekly wind and solar production Figure 2.14 in Germany...... 25 income condition and wind production. . . . 57 Pricing and clearing in the presence Figure 1.8 Hourly net load ramps for different Figure 2.15 penetration levels of solar PV...... 26 of lumpy bids ...... 61 Result with the ISO model approach. . . . 61 Figure 1.9 Wind and solar PV forecasting Figure 2.16 for Belgium...... 27 Figure 2.17 Result with the EU PX model approach. . . 61

Figure 1.10 Wind forecasting error evolution in Spain. 28 Figure 2.18 Sample of TSO’s actions to ensure the availability of reserve resources...... 64 Figure 1.11 Flexibility sources and the cost involved . .29 Operating reserves demand curve. . . . .65 Figure 1.12 System and thermal production cost Figure 2.19 for increasing solar penetration Figure 2.20 Price adder as a function of the energy in a thermal-dominated system...... 30 price for different reserve levels...... 66

Figure 1.13 Electricity production and spot prices Figure 2.21 RES balancing responsibility...... 70 November 2016 in Germany ...... 31 Figure 2.22 Standard characteristics Figure 2.1 VRE generation and the need of balancing products...... 71 for flexibility...... 35 Figure 2.23 Available ramping capacity with Figure 2.2 European power systems with and without flexibility reserve products. . 72 (or in the process of implementing) capacity mechanisms...... 37 Figure 2.24 Conventional regulation signal (blue) and the faster regulation D signal (yellow). 73 Figure 2.3 Market clearing price and market clearing volume setting in electricity auctions. . . 38 Figure 2.25 Updated benefits factor curve to optimise operations under all system conditions. . .73 Figure 2.4 Short-term market time frames: An overview ...... 41

6 LISTS OF FIGURES, TABLES AND BOXES

Figure 2.26 Annual gross aggregated electricity Figure 3.15 Process for avoiding TFO congestion production from gas-fired, solar and wind using flexibility in the distribution grid. . 128 electricity plants in the European Union. . 77 Figure 3.16 One-line diagram of PNM’s battery Figure 2.27 Coincidence of solar energy with storage pilot project...... 132 peak demand in Colorado...... 79 Figure 3.17 Framework proposed by the Italian Figure 2.28 Reference price settlement periods. . . . 86 regulator to decide upon the role of distribution companies Figure 3.1 Effect of energy storage to mitigate in storage ownership ...... 133 the impact of PV penetration on distribution costs...... 94 Figure 3.18 Network of Tesla Motor’s superchargers across North America...... 136 Figure 3.2. Expected network investments in Great Britain to accommodate Figure 3.19 Contractual relationship between low-carbon technologies connected actors involved in electric mobility . . . . 137 to the distribution network under different planning scenarios (2012-50). . .94 Figure A.1 Linked block orders...... 165

Figure 3.3 Examples of map-based information Figure A.2 Characteristics of pumped hydro for potential DG connections methodologies...... 166 in the United Kingdom...... 97

Figure 3.4 DER integration map in California LIST OF TABLES (Southern California Edison)...... 97 Table 2.1 Bid sessions for intraday market in Italy. . 43 Figure 3.5 Impact of DG on distribution voltage control strategies...... 99 Table 2.2 Typical multipart offer structure in ISO markets...... 55 Figure 3.6 Distribution of budget per country and smart grid application in Europe, Table 2.3 Energy-weighted average energy considering only demonstration and price adder (and online reserve price). . . 66 deployment projects...... 101 Table 3.1 Overview of smart grid pilots in India. . .103 Figure 3.7. Cumulative annual budget of smart Table 3.2 General guidelines for a smarter grid projects in Europe, 2002-14 . . . . .102 distribution regulation ...... 109 Figure 3.8 Categorisation of US smart grid Table 3.3 Net-metering cap at the utility demonstration projects by type . . . . .103 level in California (March 2016)...... 117 Figure 3.9. Sample daily load profile for a plastic Table 3.4 Major services that DER manufacturing company with may provide to DSOs and TSOs. . . . . 126 self-consumption from PV...... 113 Table 3.5 Current and future TSO-DSO interactions. 127 Figure 3.10 Net-metering policies across the United States in July, 2016 ...... 114 Table 3.6 Smart metering cost-benefit analysis (CBA) in Germany (left) and the United Kingdom Figure 3.11 PV production in the European Union (right)...... 130 in 2014...... 115 Table 3.7 Summarised benefit-cost analysis Figure 3.12 Possible effects of traditional regulatory of PNM’s battery storage pilot project . . .132 approach in the presence of high shares of DER ...... 116 Table 3.8 Proposed storage targets for IOUs in California (MW installed per year). . . 134 Figure 3.13 Influence of pricing signals on PV and storage investment decisions Table A.1 Bidding formats and description . . . . . 165 Figure 3.14 Treatment of prosumers’ excess energy across the United States in July, 2016. . . 118

7 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

LIST OF BOXES Box 2.20 Cross-border participation in European Capacity Remuneration Mechanisms. . . .82

Box 2.1 The Major Difference Between Box 3.1. Network investments and US and EU Market Designs...... 39 advanced solutions to integrate distribution generation...... 94 Box 2.2 Short-term electricity market timeline in Italy...... 43 Box 3.2. Regulatory mechanisms to handle a large number of DG connection Box 2.3 Intraday markets and schedule applications...... 96 adjustments ...... 44 Box 3.3. Network information disclosure Box 2.4 Intraday markets and forecast accuracy. . 50 for new DG connections...... 97 Box 2.5 Market integration Box 3.4. The need for a more active network in the European Union ...... 53 management: voltage control with DG. . .99 Box 2.6 Bidding formats in the US ISO market: Box 3.5. Non-firm distribution network Multipart offers...... 55 access as a means to prevent Box 2.7 Bidding formats network constraints ...... 100 in European power exchanges ...... 56 Box 3.6: Digitalization and cooperation between Box 2.8 Empirical evidence of the growing role energy and telecom regulators...... 105 of complex conditions and block bids in EU Box 3.7 Regulatory reforms to encourage the power exchanges: the case of Spain. . . .57 adoption of innovative grid technologies and Box 2.9 Illustrating the need for uplifts...... 59 solutions...... 107

Box 2.10 Uplifts: an increasingly relevant Box 3.8 Examples of self-consumption benefits problem in some US systems ...... 60 for commercial end users...... 113

Box 2.11 Illustrating the difference between Box 3.9 Key terms related to self-consumption the US and EU pricing and clearing. . . . .61 and net-metering...... 114

Box 2.12 TSO’s actions to ensure the availability Box 3.10 Simulation of investment decisions of reserve resources...... 64 in PV and storage under different pricing schemes...... 118 Box 2.13 Techno-economic Analysis of the Impact of Implementing the ORDC in Texas Box 3.11 Market-based earnings and new roles (Backtesting for Years 2011 and 2012). . . 66 of utilities in New York ...... 125

Box 2.14 Utility solar PV plant’s ability Box 3.12. Balancing generation and demand with to provide ancillary services to the grid. . 68 distributed flexibility...... 126

Box 2.15 Defining the characteristics of balancing Box 3.13. The benefit-cost analysis products: The European approach. . . . . 71 of a grid-connected storage system. . . .132

Box 2.16 Defining innovative products for batteries Box 3.14. Business models and drivers in the United States...... 73 for electric vehicle public charging in the United States...... 136 Box 2.17 Wind penetration in Brazilian long-term auctions ...... 79 Box 3.15 Charging service model according to the European Directive 2014/94/EU Box 2.18 Renewable participation on Alternative Fuels Infrastructure. . . . .137 in PJM’s reliability pricing model...... 80 Box A.1 The complexity of bidding some Box 2.19 Renewable participation in ISO resources with existing formats: New England’s Forward Capacity Market. . 81 the case of storage ...... 166

8 ABBREVIATIONS

ABBREVIATIONS

AEEGSI Autorità per l’Energia Elettrica ISO-NE Independent System Operator il Gas e il Sistema Idrico of New England AFID Alternative Fuels Infrastructure KWH Kilowatt hour AMI Advanced metering infrastructure M2M Machine-to-machine ANM Active network management MI Intra-day market (Italy) BAU Business-as-usual MISO Midcontinent Independent System ARRA American Recovery and Reinvest- Operator ment Act MV Medium voltage BF Benefit factor MWH Megawatt hour BRP Balancing responsible party NEG Net excess generation CAISO California Independent System NEM Net energy metering Operator NYISO New York Independent System CAPEX Capital expenditures Operator CEER Council of European Energy Regu- OFGEM Office of Gas and Electricity Markets lators OMS Outage management system CPO Charging point operator OPEX Operational expenditures CPUC California Public Utilities Commis- ORDC Operating reserves demand curve sion PJM Pennsylvania Jersey Maryland Inter- CRM Capacity remuneration mechanism connection DER Distributed energy resources PLM Peak load management DG Distributed generation PNM Public Service Company of New DLMP Distribution locational marginal price Mexico DSIRE® Database of State Incentives for PQM Power quality management Renewables and Efficiency® PV Photovoltaic DSO Distribution system operator PX Power exchange EV Electric vehicle RES Renewable energy sources EU European Union REV Reforming the energy vision FACTS Flexible Alternating Current Trans- RIIO Revenues = Incentives + Innovation mission System + Outputs FIT Feed-in tariffs RPI Retail price Index FRR Frequency restoration reserve RPM Reliability pricing model GW Gigawatts TOTEX Total expenditure ICT Information and communication TOU Time-of-use technology TSO Transmission system operator IEA International Energy Agency US United States ISGAN International Smart Grid Action VIU Vertically integrated utility Network VRE Variable renewable energy ISO Independent system operator VOLL Value of lost load

9 EXECUTIVE SUMMARY EXECUTIVE SUMMARY

BACKGROUND AND KEY MESSAGES

The world is undergoing an accelerated energy •• Distributed energy resources improve the resil- transition, one driven by environmental, techno- ience of the electricity system during extreme logical, social, and economic drivers. The ways in weather events and unscheduled technical fail- which electricity is produced, transported, and ures. consumed are essential to this transformation, •• Distributed energy resources allow consumers and will undergo profound changes over the next to take on a more active role within the power several years. These changes include the pro- system. gressive adoption of low-carbon technologies, To realise the positive contributions of renew- including through the increased deployment of ables regulation and electricity market design variable renewable energy the decentralisation of must adapt to changing conditions. The share of generation, driven by the growing availability and variable renewable energy has grown substan- reduced costs of distributed-energy resources; tially in some countries and there is an ongoing increases in regional inter-connection and mar- process of rethinking the future design of elec- ket integration; and consumer responsiveness tricity markets to make them compatible with enabled by information and communication tech- ambitious renewable-energy targets. nologies. To support the ongoing discourse, this report High shares of variable renewable energy have is primarily targeted at stakeholders from lib- been successfully integrated into the power eralised electricity markets with large shares of system in several advanced markets, at rates variable renewable energy, and provides a set well above 30% of the generation mix. Variable of policy and regulatory recommendations to renewable energy can contribute to its own inte- facilitate continued deployment. The recommen- gration in the system and bring multiple benefits: dations address different activities in the power •• Technological innovation allows variable re- supply chain from wholesale markets and sys- newable energy technologies to provide ancil- tem operations to distribution networks and end lary services, ultimately contributing to system users. Many different contexts and regulatory flexibility and reliability. frameworks are considered, although examples •• Distributed energy resources reduce energy are drawn mainly from North America and Eu- losses, peak demand and the need for network rope. The recommendations are broadly appro- reinforcements under certain conditions. priate for power systems with organisational •• Variable renewables can contribute to meet models that are similar to the ones analysed here: peak demand by providing capacity to the sys- liberalised electricity markets with high shares of tem, essentially allowing them to participate in renewables. capacity mechanisms.

11 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

These recommendations are based on sound 1.REFORMING WHOLESALE ELECTRICITY regulatory principles and reflect the complex MARKETS TO ENCOURAGE FLEXIBILITY ES inter-relationships of activities and stakeholders across the power sector. The overarching goal 1.1 Adapting short-term markets is to assist policy makers and regulators to take The presence of high shares of variable renewable timely action to facilitate the energy transition. energy increases the uncertainty in the prediction of market conditions and network constraints. The analysis is divided into two main sections. The Consequently, the time and locational granularity first part addresses the integration of variable re- of market signals should be increased with rise in newable generation within wholesale electricity variable renewable energy levels. The design of markets. It recommends reinforcing the design short-term energy markets should be enhanced of short-term markets (day-ahead and intra-day and refined at all levels, including timelines, bid- markets), balancing markets, and long-term in- ding formats, clearing and pricing rules, and their vestment signals. Adapting short-term markets integration with reserves and regulation markets. requires improving temporal and spatial granu- Some recommended reforms include: larity, increasing the details of bidding formats, and strengthening the link between energy and •• Increase time granularity by making trading reserve markets. Adapting balancing markets intervals shorter and closer to real time. As involves redefining traded products, recognising generation variability increases, market sig- the contribution of variable renewables to grid nals should be made more time-specific and stability, and avoiding dual-imbalance pricing. gate-closure times reduced to reveal the flex- Long-term mechanisms, which guide genera- ibility of resources that can respond quickly tion expansion according to the strategic view to fast-changing conditions. This can be done of government, should allow mature renewable through continuous trading, discrete intra-day technologies to compete with other generation auctions or a combination of both. technologies and to ensure that both renew- •• Increase locational granularity by using zonal able-energy support and capacity mechanisms or nodal prices. The increased deployment of are designed to minimise distortion in electricity variable renewable energy, particularly wind, markets, while taking into account the environ- may result in a constrained transmission net- mental externalities, and the need for further work. Markets must reflect such network con- renewable energy development. straints. Zonal pricing is a simplified approach to address this, but it comes at the expense of The second part of this analysis addresses the market efficiency, particularly in regional mar- regulation of distribution networks and the design kets. Zonal pricing works well when transmis- of retail tariffs to allow for smarter distribution sion congestions are structural and systematic, systems and increasingly active network users. It and when predefined price zones are easy to recommends incentivising distribution companies determine accurately. However, variable renew- to adopt a more active role in network planning able energy can alter power-flow patterns sig- and operation, and implementation of smart grids. nificantly, thus exacerbating the limitations of The report also emphasises the importance of de- zonal pricing. In this context, nodal pricing can coupling distribution remuneration from the vol- often provide better operation and investment ume of energy distributed, and shifting the goal signals but may be harder to implement. of regulation solely from investment adequacy to •• Reform wholesale-market bidding formats to also include a broader set of performance indica- incorporate increased detail in the representa- tors. The report recommends promoting self-con- tion of generation and demand characteristics. sumption through cost-reflective retail tariffs and With higher variability, and given the need to supporting advanced metering technologies. fully exploit flexibility from storage or demand Distribution companies should adopt new roles as response, it is important to rethink and reform market facilitators and distribution system oper- the bidding formats used by participants in en- ators, interacting more closely with other agents ergy markets to submit their bids. These need such as suppliers, aggregators and transmission- to go beyond simple price-quantity bids and and independent-system operators.

12 EXECUTIVE SUMMARY

move toward advanced schemes that allow explored. When service providers with different market participants to hedge against increas- response capabilities compete to provide the ingly variable short-term market conditions same balancing product, performance-based and to better represent the characteristics of remuneration can help avoid defining too many demand response and storage. reserve products and markets, and can also re- •• Adapt existing pricing and market clearing duce overall reserve requirements. rules. Current pricing and clearing rules either •• Avoiding dual-imbalance pricing. Because du- focus on minimising the cost of economic dis- al-imbalance pricing does not exactly reflect patch at the expense of having uplift payments the costs of imbalance, it distorts real-time outside the market, or on providing uniform price signals. Although portfolio aggregation payments to all market agents at the expense can mitigate the deviation risks that renewable of higher complexities. The right balance ought producers face when dual-imbalance pricing is to be met according to the policy priorities in applied, the practice still provides a competi- each context. tive advantage to larger companies over small •• Strengthen the link between energy and re- providers and distributed energy resources. serves markets. This imperative reflects an im- Where it is applied, responsibility for balancing portant key fact: supplying one of them implies supply should be assigned to each generation modifying the ability of power plants to pro- unit to prevent competitive disadvantages. vide the other. Strengthening this link involves co-optimising energy and reserve procurement 1.3. Encouraging long-term investment both in day-ahead and shorter-term energy Long-term support mechanisms, including ca- markets. Where that is difficult, frequent mar- pacity or adequacy mechanisms and support ket sessions for the procurement of reserves schemes for renewables, are widely used to guide could be organised, aligned with the timelines generation investments according to country of energy markets. Furthermore, system opera- policy priorities. When these policy interventions tors should abandon inflexible reserve require- are deemed necessary, the following guidelines ments and implement new solutions to procure should be followed: first, as far as technically and price reserves according to the actual value possible, mature renewable technologies should they provide to the system. compete with other generation technologies, even in capacity markets. Second, both renew- 1.2. Adapting balancing markets able-energy support and capacity mechanisms Redesigning of balancing markets is essen- should be designed to minimise distortions of tial for the effective and efficient integration electricity markets. of variable renewable energy. Existing rules •• Renewables should be allowed to participate should be reviewed to ensure that they reward in generation-adequacy mechanisms. Renewa- flexibility and to facilitate the effective use of ble generation can be a valuable contributor to all resources. system adequacy, especially in power systems •• Redefining balancing products. To the extent with a high share of conventional (i.e. reser- possible, balancing energy should be procured voir-based) hydropower. Therefore, regulations from the most economic resources available in should avoid introducing systematic technol- real time, even if those resources do not acquire ogy-specific market-entry barriers, and allow longer-term commitments in day-ahead or oth- renewable technologies, including variable re- er reserve markets. Furthermore, upwards and newables, to participate in generation-adequa- downwards reserves should be two distinct cy mechanisms on a level playing field, where products. and when it is technically possible. •• Facilitating variable renewable energy contri- •• Capacity mechanisms do not necessarily sub- bution to grid stability. To efficiently exploit all stitute for support mechanisms. Long-term available flexibilities, including those offered by support mechanisms should be designed in renewables, new reserve products should be a coherent way to ensure that the incentives

13 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

provided through renewable-energy support A first-come-first-served approach can mean schemes account for possible remuneration higher connection costs for later applicants due ES earned in the capacity market (as well as in to reinforcement requirements. It also results in other markets). inefficient grid development due to economies •• Economic support for renewables should be of scale. Therefore, co-ordinated approaches market compatible. Where policy makers con- should be explored such as working in batches sider offering economic support for renewable per network area. technologies, this should be done in a way that •• Disclosure of grid condition information. Infor- is compatible with markets. Several designs for mation disclosure obligations should be levied support schemes are available, all of them offer- on distribution companies such that new DG ing advantages and disadvantages. A balance units have information on the condition of the must be found between optimal investment in- grid for a point of connection. Publishing the centives and market compatibility, determined available generation-hosting capacity allows according to policy priorities. Mixed approach- DG promoters to estimate whether their applies may be explored to achieve a compromise cation will be successful and determine which solution. location will result in lower connection charges. Ultimately, this facilitates the integration of DG. 2. FOSTERING SMARTER DISTRI- •• Remunerate distribution companies based on their active grid management. Large penetra- BUTION SYSTEMS AND MORE tions of DG introduce complexities in distribu- ACTIVE NETWORK USERS tion planning since the location of DG units can 2.1. Adapting distribution systems be highly uncertain. Integrating DG efficiently requires active network management as an al- The efficient deployment of high levels of dis- ternative to conventional grid reinforcements, tributed energy resources, including distributed such as solving network constraints in real time, generation (DG), demand side management or close to it. Regulation should promote this and small-scale storage, requires innovative ap- transformation. One important way to do that proaches to planning and operating distribution is to require utilities to submit detailed business networks. Conventional grid access and connec- plans based on cost-benefit methodologies as tion rules and practices should be adapted ac- part of the remuneration process. cordingly and smart-grid technologies deployed. •• Enable advanced forms of contracting be- •• Rethinking planning for DG. Grid connection tween distribution companies, generators has traditionally followed a “fit-and-forget” and consumers. Active network management approach, i.e. reinforcing the grid as much as requires the development of smarter grids as necessary to prevent any operational problems. well as closer interaction between all relevant This is a safe and robust strategy and requires actors. The latter can be achieved through flex- very low levels of network monitoring. But as ible connection contracts that limit curtailment DG penetration levels increase such an ap- of generation or demand in exchange for some proach can be costly, especially in areas with form of compensation or under specific condi- high concentrations of DG, and can cause long tions. As the presence of DER grows, more ad- lead times for connecting new DG sources. vanced forms of contracting flexibility services, Therefore, regulators should gradually abandon such as bilateral agreements or market-based the “fit-and-forget” approach as DG penetra- approaches, could be implemented. tion levels grow, and rethink network planning •• Promote smart grids. Technology risks, and the and grid connection. absence of economic incentives, prevent the •• Co-ordinated approach to grid connection. development of smarter distribution grids. Pol- Grid-connection application processes should icy and regulation should promote and support be reviewed to speed up DG connection and innovation, implementation of pilot projects, to allocate grid capacity more efficiently. the exchange of lessons learned, and the shar-

14 EXECUTIVE SUMMARY

ing of best practices. The creation of public-pri- Rethinking the remuneration of distribution vate collaborative networks and the definition companies. As the penetration levels of distribut- of knowledge sharing and information disclo- ed energy resources increases, greater changes sure obligations can facilitate information ex- to current regulatory approaches are required. change. The regulation focus should shift from short-term cost reductions to the promotion of long-term 2.2. Promoting long-term efficiency efficiency. Distribution companies should also be Distribution regulation has conventionally focused encouraged to implement innovative grid plan- on promoting adequate investment levels and ning and operation solutions (i.e. smarter distri- short-term efficiency gains, while controlling the bution grids). This involves: quality of service. This approach works well with •• The focus of regulation should shift from stable and predictable grid technologies and net- ensuring that companies invest sufficient- work users. However, conventional remuneration ly in networks to assessing grid operators formulas and cost-assessment methodologies based on their performance, as measured are less optimal as the presence of distributed by an extended set of indicators. Those in- generation increases. With moderate distributed dicators could include customer satisfac- energy resource penetration levels, the poten- tion, grid-connection lead times, the carbon tially negative effects of distributed generation footprint or available distributed generation can be mitigated with a few regulatory tweaks. hosting capacity. When these indicators can Nevertheless, efficiently integrating high shares be objectively measured and controlled by of distributed energy resources requires a major the distribution companies, incentive (and regulatory overhaul. penalty) mechanisms can be implemented. Decoupling distribution volumes from energy •• The required novel methods of managing the volumes. Cost-assessment methods generally network reduce the need for investment in dis- applied to determine allowed revenues of distribution assets but increase operational ex- tribution companies are usually unable to accu- penditures. Traditional regulatory approaches rately quantify the impact of distributed energy discourage novel methods if distribution com- resources on network costs. To compound this panies are remunerated mostly based on their situation, distribution remuneration usually capital investment. Incentive systems should depends on the volume of energy distribut- reward both operational and capital expendi- ed, which DG (and some forms of demand side tures of distribution companies. management) dminishes. As a result, distributed •• Regulatory benchmarking, a method of de- generation, self-consumption or energy-efficient termining remuneration for distribution com- measures may translate into a reduction in rev- panies, usually relies on past information. It enues without a corresponding drop in costs. implicitly assumes that future developments To mitigate such a negative impact, distribution will follow a similar trend. This assumption is revenues should be independent of the volume questionable in a context requiring innova- of energy distributed. This decoupling of revenue tion. Hence, cost-assessment methodologies essentially consists of adjusting network tariffs should increasingly rely on forecasted data ex post so that distribution companies recoup ex- and well-justified investment plans submit- actly the allowed revenues. Moreover, if cost-as- ted by the regulated companies. sessment tools are unable to capture the impact •• Promoting efficient investment in distribution of DG on distribution costs, economic compensa- networks requires adopting a long-term per- tion on top of conventional revenue allowances spective given the long life of the assets and may be necessary to account for it. the time required before innovation yields benefits. Regulators should progressively extend the length of the regulatory period to incentivise distribution companies to pursue

15 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

long-term efficiency. Remuneration formu- of self-consumption and to stimulate end users’ las should be more flexible to accommodate demand response, including distributed storage. ES possible uncertainties in such a dynamic en- Economies of scale and standardisation are im- vironment. This may include the introduction portant when deploying advanced meters. of profit-sharing schemes and automatic-ad- •• The changes that advanced power systems are justment factors. experiencing are strictly interlinked with the power system digitalization. The evolving role 2.3. Improving tariffs and metering of information and communication technolo- Self-consumption and the adoption of distributed gies for the efficient management of the power storage behind the meter can yield benefits for system calls therefore for a close cooperation both end-users and the power system as a whole. between electricity and telecommunication Therefore, regulation should actively promote regulators. self-consumption by adopting a cost-reflective design for retail tariffs and supporting the roll out 2.4. Encouraging the new roles of advanced metering technologies. of distribution companies As the energy transition evolves, a growing share •• High levels of self-consumption may negative- of the resources needed to ensure secure and ly impact the financial viability of distribution flexible system operations will be connected at utilities and the recovery of fixed power-system the distribution level. In this new environment, costs. Such issues become more prevalent in distribution companies must bridge the gap be- markets with high shares of renewable sources tween flexibility providers (i.e., distributed gen- supported by net-metering policies since they erators, responsive demand and aggregators), implicitly value the energy injected into the markets and transmission/independent system grid at the retail electricity price. To keep up operators. To do this, they should adapt their with these developments, regulations in many planning and operational practices accordingly systems have limited individual or aggregate and play new roles as market facilitators and dis- installed capacity, reduced the period of time tribution system operators. over which energy injections can offset energy withdrawals, and changed the structure of retail •• Regulation should allow distributed energy re- tariffs and compensation rules. However, these sources to participate in upstream energy and do not provide a real long-term solution for juris- ancillary services, particularly when these re- dictions with high penetration of active agents. sources become widespread. Distribution com- •• The sustainable development of high levels of panies should facilitate this participation and on-site generation in mature liberalised markets carry out activities such as ex ante technical entails adoption of self-consumption schemes validation, to ensure that no constraints arise in with hourly netting intervals, or even shorter. In the distribution grid, and ex post verification of addition, retail tariffs should be cost reflective. the provision of the services. They should be based on the value of electricity •• To facilitate well-functioning retail markets and at each time and location, the individual contri- the participation of distributed energy resourc- bution of the network users to network costs, es in wholesale markets, it is critical that mar- and a charge to recover other regulated costs ket agents have transparent and non-discrim- so that the economic signals sent by energy inatory access to metering data. This might and network charges are not distorted. be seen as a conventional task of distribution •• Advanced-metering infrastructure should be companies, but concerns arise when a metering installed so that adequate locational and time data manager, traditionally a local distribution granularity in the tariffs can be communicated company, is also a market participant. In this to end consumers. Electronic meters capable context, alternative models for data manage- of recording bidirectional energy flows every ment could be explored, such as creating a new few minutes are needed for the development regulated entity responsible for data manage-

16 EXECUTIVE SUMMARY

ment (central hub) or opting for a decentralised charging points as part of the regulated asset approach. There is no consensus on the most base may imply that rate payers would be sub- appropriate model, but regulations must al- sidising EV users. To avoid such problems, other ways ensure non-discriminatory access to data policy alternatives might be adopted to provide and protect consumers’ privacy, particularly the initial policy push. after the deployment of advanced metering. •• Distributed storage will be another game •• Distribution companies should make use of dis- changer in the power sector, also for its po- tributed energy resource flexibilities by active- tential to supply grid-support services. For ly integrating with the resources connected to this reason, distribution companies may seek their grids. Ad hoc regulatory mechanisms such to own and operate storage devices. However, as non-firm connection agreements, bilateral unbundling provisions could rule this possibility agreements or local markets may be necessary. out since storage operators may wish to pro- Regulators should clearly define the respon- vide other services under competition to obtain sibilities of distribution companies, especially a positive business case. Thus, exemptions on where a distribution company belongs to a ver- the unbundling obligations may be considered tically-integrated company in a context of retail in some cases. In others, distribution compa- competition. nies may be entitled to contract services with storage operators through auctioning. 2.5. Facilitating the development of infrastructure for storage and electric vehicles The development of electric mobility requires careful regulation of the contractual relationship between the various actors involved: electricity distribution operators, electricity suppliers, charging point operators, mobility service providers, and electric vehicle (EV) drivers. Distribution companies will play a key role in the deployment and operation of new grid-edge infrastructure such as public EV charging stations, or distributed storage. The major regulatory question is whether to consider them part of the business model of distribution companies or open them to competition. The former can collide with unbundling rules and lead to a suboptimal utilisation of these technologies, while the latter may make it harder for distribution companies to benefit from their potential contribution to grid planning and operation.

•• Market forces alone may not be able to foster the development of public charging infrastructure. Policy makers may have to kick-start the infrastructure development, for example by giving distribution companies responsibilities. However, this may be challenging. On the one hand, unbundling rules may prevent distribution companies from selling electricity to electric-vehicle users; on the other hand, treating EV

17 POWER SECTOR TRANSITION POWER SECTOR TRANSITION 1

1.1 THE ENERGY TRANSITION AND and the building and industry sectors must use ITS IMPLICATIONS FOR THE POWER more electricity in their heating and cooling pro- SECTOR cesses. The global energy transition is underway, driv- The increased electrification of end energy uses en by ambitions to improve energy security, ad- will take place in the midst of several other ma- dress environmental impacts, and achieve uni- jor trends that are also shaping the future of the versal energy access. The transition will involve power sector, including: 1) the integration of pow- a fundamental rethinking of the way energy is er systems into larger entities at regional or su- produced, distributed and consumed which will pranational levels, seeking more efficiency and bring changes to the way societies operate and a reliability; 2) the increased interplay between wide range economy-wide benefits. The political the power sector and adjacent ones, such as gas, commitment to the energy transition is reflected and information and communication technology; in the adoption of the Sustainable Development 3) the spectacular cost reduction and technology Goals, and the ratification of the Paris Agreement. improvement in promising technologies, which could become disruptive, especially electricity The adoption of renewable energy technolo- storage; and 4) the increasing presence of dis- gies is a key pillar of the energy transition. Over tributed energy resources (DER), which include the past decade, as the technologies have ma- distributed generation, demand response, dis- tured and costs decreased (Figure 1.1), their de- tributed storage and other devices. ployment has grown at a tremendous pace in the power sector. In fact, for the past five years, re- Energy storage is viewed to be the next game newables capacity additions have exceeded ad- changer in the power sector. IRENA forecasts ditions of conventional power capacity. Looking show that the cost of energy services from bat- forward, mainly because of the potential of so- tery electricity storage will continue to decline lar and wind technologies, the power sector will due to reductions in the installed costs and im- play an important role in the energy transition. provements in battery technology performance To achieve climate targets, part of the transpor- (Figure 1.2). tation sector has to switch to electrical mobility,

19 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 1.1 Levelised cost of electricity for utility-scale power (ranges and averages), 2010 and 2016

1 400 350

Weighted

a 250 average

200

USD/MWh 150

100 Fossil fuel cost range 50

0 2010 2016 2010 2016 2010 2016 2010 2016 2010 2016 2010 2016 2010 2016 Biomass Geothermal Hydropower Solar PV Solar Thermal Oshore Onshore Wind Wind

Note: a) MWh: megawatt-hour. b) All costs are in 2016 USD. Weighted Average Cost of Capital is 7.5% for OECD and China and 10% for Rest of World. Source: IRENA, 2017a

According to most scenarios, future investment a crucial role in the transition to a more decen- in power generation will be dominated by RES tralised system, thereby shaping the electric grid worldwide. Among RES investment, wind and of the future. For example, advanced meters can solar technologies would represent the highest allow real-time pricing and fully automated bill- shares. In particular, the use of decentralised so- ing. With over 12 million smart meters, California lar photovoltaics (PV) is expected to rise sharp- has reached close to 100% penetration, enabling ly, due to the significant cost reductions and the consumers and utilities to access added function- support of policies. The technology has already alities. reached significant penetration levels in some The European Union also has an ambitious pro- cases. In California, for instance, around 2 GW gram for deployment of smart meters in mem- of distributed solar PV has been installed. Figure ber states before 2020. Italy pioneered a com- 1.3 shows the growth of installed solar PV capac- plete installation of advanced meters more than ity between 2007 and 2015 under the California a decade ago. Of course, what matters is not the Solar Incentive program and the corresponding physical existence of these meters, but taking full observed reduction in average installation cost advantage of their potential. Concerns related to during this period. the availability of abundant online information Changing role for consumers and utilities about each individual network user, such as pri- −− vacy, must also be addressed. The transition is not driven by cost reduction alone. Another relevant factor is the more active Under this new paradigm of decentralised re- role of consumers, enabled by the possibility of sources and customer engagement, new busi- generating their own electricity and by the de- ness models and commercial strategies are ployment of new technologies, such as advanced emerging. With the rise of distributed generation, meters or load automation systems. Better in- individuals and communities have greater con- formed and more responsive consumers will play trol on generation and consumption of energy.

20 POWER SECTOR TRANSITION

Figure 1.2 Energy installation costs and cycle life of battery storage technologies, 2016 and 2030

2030 2030 Titanate ZnBr Flow

12,000

10,000 2016 2016 Titanate ZnBr Flow 2030 8,000 2030 NaS LiFePO4 2030 6,000 NMC/LMO 2016 NaS 4,000 2030 2030 NCA VRLA Cycle life (equivalent full-cycles) (equivalent life Cycle 2030 2,000 2016 2016 Flooded LA 2016 LiFePO4 NMC/LMO 2016 VRLA 2016 NCA Flooded LA 0 1,100 1,000 900 800 700 600 500 400 300 200 100 0 Energy installation cost (USD/kWh)

Flooded LA VRLA NCA NaS NMC/LMO LiFePO4 ZnBr Flow Titanate

Note: Titanate: Lithium Titanate; ZnBr Flow: Zinc-Bromine; NaS: Sodium Sulphur; NMC/LMO - Nickel Manganese Cobalt Oxide (NMC) / Lithium Manganese Oxide (LMO); LiFePO4: Lithium Iron Phosphate; NCA: Lithium Nickel Cobalt Aluminium; VRLA: Valve regulated lead-acid; Flooded LA: Flooded lead-acid Source: IRENA, 2017b

For example, 35% of Germany’s RES installation is −− Power system integration owned by citizens, whereas the aggregated share Together with the structural and ownership trans- in total distributed generation capacity of the big formations associated with the rising presence of four incumbent utilities, namely E.ON, RWE, Vat- decentralised resources, other important drivers tenfall and EnBW, is only 5% (BEE, 2014). of change can be observed in the power sector. The increasing share of RES is affecting the capacity factor of fossil fuel power plants. Accord- Figure 1.3 Solar PV installed capacity within Califor- ingly, several incumbent utilities changed their nia Solar Incentive (CSI) program against the evolu- roles and strategies for their energy business- tion of average cost of solar PV capacity es. In September 2013, RWE, Germany’s largest power producer, decided to depart radically from 12 1,800 its traditional business model based on large- 1,600 10 scale thermal power production and to become 1,400 a service company, increasingly acting as proj- 8 1,200 1,000 6 ect enabler, operator and system integrator of re- 800 newables. Similarly, at the end of 2014, E.ON an- 4 600

Average cost ( $ /W) cost Average 400 nounced that it is spinning off conventional power 2 200 CSI installed capacity (MW) capacity CSI installed plants to focus on RES, distribution network and 0 0 customer solutions. RWE and E.ON illustrate the 2007 2008 2009 2010 2011 2012 2013 2014 2015 strategies of traditional industry to adapt to the Source: California Solar Statistics, 2016 power sector transformation.

21 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

The integration of power systems into increasing- the long-term financial viability of the power sys- ly large markets is one. Power pools encompass- tem. Although several countries have successfully 1 ing several electric utilities under traditional reg- incorporated a large share of variable renewable ulation were created in the United States during energy (VRE), over 30% of total electricity gener- the 1970s, and they have evolved into the present ation, without compromising the reliability of the regional markets, known as regional transmission electricity supply, achieving the energy transition organisations and run by independent system involves changes at many levels. operators. One of these is the adaptation of liberalized elec- In the European context, electricity and gas mar- tricity markets to efficiently accommodate even kets form a pillar of the regional energy policy. higher shares of VRE. European electricity market liberalisation, which began in the 1990s, represents one of the most 1.2 Power sector organisation extensive cross-jurisdiction reform of the elec- This section briefly describes the main activities tricity sector involving integration of distinct in the power supply chain and discusses the dif- state-level or national electricity markets (CEEPR, ferent possible organisational approaches that 2005). may be found worldwide. Today, national and regional day-ahead, intraday Broadly speaking, the electricity supply chain can and balancing markets are under reform to cre- be divided into four distinct activities: genera- ate single platforms where all European market tion, transmission, distribution and retailing. Un- players can trade the different energy products, til recently, power generation has usually taken subject only to the limitations of cross-border in- place in large centralised power plants located far terconnection capacities.1 away from the load centers. Therefore, transmission grids are necessary to transport electricity Since the early 1990s, Latin American countries over long distances. In the downstream segment have launched diverse initiatives to increase in- of the power system, distribution networks sup- terconnection capacities and to create integrat- ply electricity locally from the transmission sub- ed regional electricity markets. Mercado Eléctrico stations to end consumers. The traditional role of Regional, the Central American electricity market distribution companies as network operators and (MER), is a present reality, having formally be- suppliers of electricity to consumers is being re- gun operations in 2013. Another initiative is the visited because of the need to integrate growing Interconnected Andean Electricity System SINEA levels of distributed generation (DG) and to fa- which is in the process of creating a sub-regional cilitate retail market competition and consumers’ electricity market, including Colombia, Ecuador, demand response. Lately, some countries are lib- Peru and Bolivia. eralising the retail segment with the goal of pro- Other examples are the well-established Aus- moting efficiency. tralian National Electricity Market, the six inde- The traditional boundaries between transmission pendently operated regional markets in China and distribution systems were largely driven by and the four power pools in Sub-Saharan Africa, the different functions these networks fulfilled which are still under establishment. within the power system. On the one hand, trans- The decentralisation and growing penetration of mission grids have conventionally had a decisive RES are bound to transform the power sector. impact on the system generation dispatch and its While some challenges remain, favorable policies efficiency, as most generation was connected to and regulations can bridge the gap and avoid them. On the other hand, the main role of distri- technology lock-ins that prevent the achievement bution networks has traditionally been to bring of a suitable low carbon path. These policies in- the electricity to the end consumer while ensuring clude those that reduce barriers, promote invest- adequate levels of continuity of supply. However, ments, create a fair level playing field and ensure these boundaries, both at the network (physical

22 1. http://www.acer.europa.eu/electricity/regional_initiatives/Pages/default.aspx POWER SECTOR TRANSITION

asset) and system operation (their use) levels, are made by the VIUs under the approval of the cor- evolving with the progressive decentralisation of responding agency, considering both generation dispatchable resources (DG, demand response, and transmission network expansion as a whole, distributed storage). whereas medium-term and short-term operational decisions are made on the basis of cost minimi- Furthermore, power systems need to be man- sation and a centralised cost-based dispatch. This aged as a whole so that security of supply is en- power sector structure is illustrated in Figure 1.4. sured over both the short and long term. Pow- er system planning and operation comprise a chain of sequential decisions, ranging from gen- Figure 1.4 Traditional power sector organisation eration capacity expansion and transmission net- with a vertically integrated utility in work planning in the long term to short-term de- each territorial franchise cisions such as the dispatch of generation units and the utilisation of reserves. Besides technical constraints, these decisions must be guided by economic considerations so that social welfare is Generation maximised, i.e., this is a cost minimisation subject to prescribed reliability targets. The main differences in power system management lie in how the responsibilities of the different decisions are Transmission allocated among stakeholders. System −− Different power sector structures Operation A salient feature of the energy transition will be Distribution the integration of growing shares of RES into the power system. The policy and regulatory adap- Retailing tations that are required to achieve this affect all the segments of the electricity supply chain and Note: Each colour represents a vertically integrated utility vary depending on the existing structure of the power sector. The essential differences among these alternatives correspond to the allocation Some countries have opened up their power sys- of decision-making among stakeholders and the tems to allow new entrants – known as indepen- degree of liberalisation and restructuring intro- dent power producers (IPPs) – in the generation duced in the segments of the electricity chain segment. This change in paradigm may be moti- (Batlle and Ocaña, 2013). vated in response to the difficulties faced by publicly owned incumbents to finance new invest- The traditional power sector paradigm relies on ments in generation and/or by the will to make centralised decision-making by vertically inte- possible the development of RES or combined grated utilities (VIUs), owning both generation heat and power facilities. The contracts signed and network assets, with the obligation to supply by incumbents with IPPs, allow them to recover end consumers within their territorial franchises. their generation costs (fixed and variable) and in- In this context, the state has a predominant role clude provisions concerning their duties and ob- since VIUs are either publicly owned companies, ligations (e.g., reliability level, ancillary services, or privately owned firms subject to some sort audits, etc.). of cost regulation and supervision by a govern- mental agency. In some countries, privately and In the markets where restructuring and liberali- publicly owned companies coexist and are both sation has taken place, it usually started by in- subject to the same traditional cost-of-service troducing competition into the generation sector regulation. Investment decisions are centrally at the wholesale level. This implies allowing new

23 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

entrants to compete with the previous incum- at the retail level. In some systems, the right to bent generation company, oftentimes a national freely choose a supplier has been established for 1 company owning strategic resources (e.g., hydro- all consumers at once; while in others, it has been power), or splitting up the previous incumbent progressively introduced, starting from the larg- generation firm into a few smaller generation est consumers to the smallest residential ones. companies allocated to private investors through Regulated default tariffs are frequently made auctioning processes. These generation compa- available for those consumers who do not wish to nies then sell their production through the inter- shop around for a supplier or who are temporarily mediation of a market operator via some sort of without a contract. spot-trading platform, where participation is fre- Lastly, some countries have decided to go be- quently mandatory, and the resulting wholesale yond the model based on a single compulsory electricity prices are passed through to consum- market operator or an ISO by allowing the agents ers by the distribution companies (see Figure 1.5). to establish bilateral contracts with one another. The function of operating the market can either be Thus, generators and retailers (or large consum- carried out by an entity independent of the sys- ers) may freely sign bilateral agreements, with tem operator (transmission system operator plus the organised spot markets simply being another market operator model), or by a separate market alternative. This model is depicted in Figure 1.6. operator. In the former circumstance, the agent in charge of system and market operation is usually 1.3 Renewable energy impact on the referred to as an Independent System Operator power sector operation (ISO). Contrasting the need to centralise the op- This section describes the main reasons why a eration of the system and the network expansion large penetration of VRE imposes the need to under a single entity, under the supervision of the adapt the design of power markets. While the im- regulatory authority, the transmission grid is fre- pact that high shares of these sources have on quently owned by one or more different agents. the operation of electricity systems has already Some countries that have liberalised their elec- been assessed in the literature (e.g. NREL 2013a; tricity markets have also introduced competition Holttinen et al., 2013), international experiences

Figure 1.5 Competition wholesale market and regu- Figure 1.6 Fully liberalised power sector with com- lated distribution companies acting also as petition at wholesale and retail level with retailers bilateral contracts

Generation Generation

Transmission Transmission

System Market Operator ISO Operation System Operation

Distribution Distribution

Retailing Retailing

24 POWER SECTOR TRANSITION

can illustrate the particular operational charac- complementarity (due to inverse correlation) be- teristics of wind and solar, and the way they can tween wind and solar production. affect both short-term operation and long-term We shall distinguish two different characteristics planning in liberalised contexts. of VRE that present challenges for their integra- 1.3.1 System impacts of variable tion: variability and limited resource predictabil- renewable energy ity. These characteristics make solar and wind completely different resources when it comes to VRE technologies, in particular wind and solar, market integration. present unique characteristics that have significant impacts on the power system. −− Variability Seasonality and complementarity The challenge associated with variable resourc- −− es is the increased need for flexibility, including The variability patterns of wind and solar heav- ramping requirements to keep the system bal- ily depend on the climatological characteristics anced depending on resource availability. Figure of each system. The assessment of the season- 1.8. shows the hourly ramps (in GW/hour) that ality in variable energy resources production al- have to be met in the Electric Reliability Coun- lows discovering correlations among different re- cil of Texas (ERCOT) system under different solar sources, as well as with demand. In Germany, for penetration scenarios. Solar production is useful instance, peak demand takes place during winter, to reduce morning ramp needs since it is well cor- mainly due to the demand corresponding to elec- related with the morning demand peak, however, tric space heating. Figure 1.7 depicts the annual a large amount of solar production shifts the net variation of wind and solar PV production in Ger- load peak and steep ramps to the evening. This many. While solar PV production is greater during highlights the importance of sending accurate summer, wind production is higher during win- (i.e. time sensitive) price signals that can drive inter. Consequently, wind production is positively vestment to efficiently integrate high penetration correlated with demand, but there is also certain levels of renewables.

Figure 1.7 Weekly wind and solar production in Germany

Weekly Production: Solar year 2014 1.5 1.0 TWh 0.5

1 4 8 12 16 20 24 28 32 36 40 44 48 52

Weekly Production: Wind year 2014

2.5 2.0 1.5 TWh 1.0 0.5

1 4 8 12 16 20 24 28 32 36 40 44 48 52

Source: Burger, 2014

25 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 1.8 Hourly net load ramps for different penetration levels of solar PV

1 8

GW/h 0

-2

-4

-6 5 10 15 20 25 Hours

No PV PV 6GW PV 18GW PV 36GW

Source: MIT, 2015

−− Predictability tions close to the demand could offset demand Another characteristic is related to resource pre- growth, reduce peak load and allow distribution dictability. Wind may be relatively unpredictable, companies to defer grid enforcements. However, as shown in Figure 1.9 which shows an example RES not located near the demand could lead to from Belgium. This example depicts the week- the need for further investments in transmission ahead and day-ahead forecasts versus the actu- networks and sometimes cause network conges- al measured production. The forecast accuracy tion. This highlights the need to reinforce the grid improves when it is made closer to real time. As as well as to send location-specific price signals in seen in Figure 1.9 also shows that PV forecasting liberalised markets. is more accurate than wind forecasting. 1.3.2 Market impacts of variable Much progress has been made in recent years on renewable energy wind forecasting tools; Figure 1.10 shows the ex- The impacts of a large penetration of VRE on lib- perience of the Spanish power system, for exam- eralised power systems, and specifically on pro- ple, where the accuracy of wind forecasts has dra- duction costs and market prices, heavily depend matically increased since 2008. on the characteristics of the generation mix. In particular whether the generation mix is energy- Geographical dispersion −− or capacity-constrained; and its degree of flexi- Another relevant characteristic of RES, which bility. makes them different from conventional thermal generation, is that the generation assets are located where abundant resources are available, sometimes leading to dispersed installations or to generation assets being relatively distant from demand centres. In specific cases and if the right regulations are in place, dispersed installa-

26 POWER SECTOR TRANSITION

Figure 1.9 Wind and solar PV forecasting for Belgium

1,000

750

500 MW

250

0 1/09 2/09 3/09 4/09 5/09 6/09 7/09 8/09 9/09 Time Horizon Wind forecasting Measured & Upscaled Day-ahead forecast Week-ahead forecast

2,500

2,000

1,500

1,000 MW 500

0 8/07 9/07 10/07 11/07 12/07 13/07 14/07 15/07 Time Horizon Solar PV forecasting Measured & Upscaled Intraday forecast Day-ahead Forecast

Source: ELIA, 2016

Capacity-constrained versus energy- Energy-constrained systems, on the contrary, are constrained systems (system adequacy) characterised by a value of peak demand sig- In capacity-constrained systems, scarcity prob- nificantly lower than installed capacity, and only lems arise when there is not enough installed constrained when the system runs out of avail- capacity available (MW) to satisfy demand at a able energy. These are typically hydro-dominat- given moment due to, for instance, forced out- ed systems with large reservoir capacity, but age of thermal plants. These systems cannot sat- they also include gas-constrained systems (e.g., isfy the peak demand but could certainly have subject to potential gas transmission constraints. enough energy available to satisfy demand on a These systems can meet the peak demand but given day. The largest portion of the capacity mix the challenge in these systems is to supply the to- in these systems is usually conventional thermal tal amount of energy required (in megawatt-hour generation. terms). These represent the two extremes, but systems can also be both capacity- and energy- constrained at the same time.

27 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 1.10 Wind forecasting error evolution in Spain

1 25

5 % Mean Absolute Error / Average production / Average Error % Mean Absolute

0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 Predictions horizon, hours 2008 2009 2010 2011 2012 2013 2014 2015

Source: REE, 2016

Flexible versus inflexible systems is called the merit order effect, which tends to de- (system operation) crease wholesale prices. A flexible system, meaning a system where gen- In inflexible thermal-dominated systems, signifi- eration can deal both with short-term variability cant penetration of VRE (especially PV) increases and limited predictability with a moderate impact the cycling requirements of conventional thermal on production costs, is due to the availability of plants. The change in the net load shape will force different possible resources. Figure 1.11 shows a plants to change their output more often to meet qualitative representation of the cost associated ramps of net demand, and to start up and shut with different sources of flexibility. down frequently. This context is illustrated in Fig- An inflexible generation mix involves a significant ure 1.12, from a simulation of different solar pen- increase in costs in the short term when genera- etration levels in the ERCOT system. Initially solar tion needs to follow a variable and unpredictable production can reduce total short-term system net load shape (i.e. demand minus VRE genera- costs and thermal generation costs (fuel costs), tion). This is typically the case of systems rely- but as solar penetration increases, thermal units ing on old thermal plants originally envisioned operation is affected by the cycling, thus increas- to produce base-load. Again, systems can be in- ing its cost. between these two ‘extremes’. Therefore, the total thermal cost will generally decrease (for its associated production decreases), Impact on spot prices −− but the production cost of thermal plants gener- The short-term impact of VRE on prices is larg- ating, and therefore setting the market prices in er in inflexible, thermal-dominated systems. Vari- some hours, may increase. This can translate into able renewables that very low or even zero vari- two impacts on energy prices: the first effect leads able costs tend to displace the most expensive to lower average prices; the second effect may be variable cost units, thus changing the marginal sharper and higher prices during a reduced num- price of the generated power of the system. This ber of hours because of increased cycling.

28 POWER SECTOR TRANSITION

Figure 1.11 Flexibility sources and the cost involved

Involuntary Load Chemical Storage Shedding

Transmission Expansion

Coal Ramping Residential Transmission Demand Response CT and CCGT Reinforcement Pumped Hydro Gas Ramping Storage Strategic RE Curtailment Cost Advanced Network Thermal Storage Expanded Balancing Joint Market Operation Management Footprint/Joint Hydro Ramping Industrial & System Operation Increased Ancillary Commercial Demand Service Liquidity Sub-hourly Response Scheduling and Improved Energy Dispatch Market Design RE Forecasting Option costs are system-dependent

Grid Codes and evolving over time

SYSTEM FLEXIBLE MARKETS LOAD NETWORKS STORAGE OPERATION GENERATION

Type of Intervention

Source: Miller, 2015

29 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 1.12 System and thermal production cost for increasing solar penetration in a thermal-dominated system

1 Total Short-term Production Cost Average Thermal Producion Cost 10.2 10.0 41 9.8 Merit Order Eect Cycling 9.6 40 9.4 9.2 39 9.0 USD/MWh Billion USD 8.8 38 8.6 8.4 37 8.2 0 6 12 18 24 30 36 42 0 6 12 18 24 30 36 42 Solar PV penetration Solar PV penetration (% peak demand) (% peak demand)

Source: MIT, 2015

The impact that increasing VRE penetration will put below a minimum technical output may cause have on short-term prices can be mitigated in flex- negative prices to appear. This problem is more ible, energy-constrained systems. Hydropower, acute in inflexible, capacity-constrained power for instance, can easily provide the much-needed systems (IADB, 2014). These issues will be ana- flexibility to the system and contribute to absorb lysed in detail in chapter 2, which provides rec- the steep ramps in the net load. The same hap- ommendations on how to adapt the wholesale pens if there are other forms of flexibility, as dis- market design for the efficient integration of high cussed in the previous section. level of VRE. −− Impact on price convergence and volatility 1.4 Conclusions As mentioned above, in some cases the output This chapter has outlined the fundamental role re- of VRE may be difficult to predict. Large devia- newable energy technologies will play in achiev- tions in day-ahead and intraday forecasts cause ing the energy transition. The power sector is at a divergence between day-ahead and intraday the forefront of the transition with renewables ca- prices. In some cases, this price divergence can pacity growing at a tremendous pace, outpacing be very large, as in the example of Figure 1.13, those of conventional energy sources. The grow- where an unexpected increase in wind availabil- ing deployment of VRE technologies and rise in ity caused a sudden drop in electricity prices in distributed generation are some of the factors intraday markets. Furthermore, intertemporal that are driving the transition in the power sec- production constraints coupled with the inability tor. VRE already accounts for a large share of the of many thermal generators to reduce their out- electricity mix in several markets. Further accelerating the transition, while ensuring high standards of economic efficiency and technical reliability, requires supportive policy and regulatory measures.

30 POWER SECTOR TRANSITION

Figure 1.13 Electricity production and spot prices, November 2016 in Germany

80.00 200.00

70.00 175.00

60.00 150.00

50.00 125.00 Price (Euro/MWh)

40.00 100.00

30.00 75.00

20.00 50.00 Power (GW) Power

10.00 25.00

0.00 0.00

-10.00 -25.00

-20.00 -500 01.11 15.11. 17.11. 20.11. 20.11. 24.11. 26.11. 29:11. 30.11. 09:53 17:26 01:00 08:33 16:06 23:40 07:13 23:00 Date

Import Balance Conventional > 100 MW Wind Solar Day Ahead Auction Intraday Continuous (right axis) Average Price (right axis)

Source: Fraunhofer ISE, 2016

For the sustainable development and greater renewable energy deployment, adequate measures are needed to adapt the market design to address system level impacts associated with variability, resource predictability and geographical dispersion, and market level impacts that include effects on spot prices. Indeed, embracing and supporting the energy transition will bring a wide range of economy-wide benefits, in addition to positive impacts on the environment and energy security. The remainder of this report mainly focuses on adapting policy and regulations for two elements of an advanced liberalised power system – wholesale market and distribution networks. It analyses the key impacts of variable and distributed renewables for each element and provides recommendations for policy makers and regulators in relevant markets.

31 WHOLESALE MARKET DESIGN WHOLESALE MARKET DESIGN WHOLESALE MARKET 2 DESIGN

2.1. INTRODUCTION As a consequence, and as reviewed in Chapter 1, the original configuration and functioning of pow- The electricity market designs implemented to er systems worldwide is changing, and will con- date, with their strengths and weaknesses, are tinue to change in the near future. the results of gradual adaptations to the physical characteristics of each particular power system. All these transformations are taking place at a Although no two market designs are alike – for no very fast pace. Adapting market designs to re- two power systems are completely alike – elec- alize the maximum benefits of existing and new tricity markets have been traditionally tailored to resources is a formidable challenge. a context in which grid-connected, large-scale This chapter reviews several important issues re- power plants supplied electricity in response to a lated to the design of electricity markets, issues relatively easy-to-forecast, passive demand. Until that need to be reconsidered in today’s new set- recently, electricity demand has been assumed to ting. In doing so, it analyses three time frames: 1) be highly inelastic and mainly dependent on tem- long term; 2) day ahead and intraday and 3) very perature and expected economic activity. short term (close to real time). The reconception of power systems −− 2.1.1. Electricity markets, products is recasting wholesale market designs and time frames Before the power industry reached a consensus Electricity markets comprise all the commercial on how to most efficiently implement electricity transactions involving energy as well as other markets, new driving factors for change put cur- complementary products related to the supply of rent market designs into question. Worldwide, electricity (e.g., ancillary services, reliability prod- electricity systems are experiencing one of the ucts traded in capacity mechanisms, etc.).1 most profound transformations in their history. To deepen the current transition, power systems Procuring products aside from energy is essential need to integrate large amounts of variable re- to ensure that the delivery of electricity is suffi- newable generation resources (often distributed). ciently secure and of adequate quality. The num-

1. This general overview is based on Batlle (2013).

33 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

ber of these complementary products and their •• In the day-ahead market – covered in Section exact definition vary from one system to another. 2.2 – trading takes place one day before the 2 Notwithstanding, energy and ancillary services delivery of electricity. Market members sub- are jointly produced; they are not typically jointly mit their bids and offers, and a clearing algo- procured, as energy procurement is left to mar- rithm determines the market price for each ket participants whereas the system operator is settlement period (e.g., for each hour) of the in charge of ancillary services’ procurement. As following day and the cleared quantities. It is we shall see later, there are alternative ways to in this time frame where the largest fraction of integrate ancillary services and energy procure- operating reserves is often acquired. After the ment. day-ahead, and before real-time delivery of electricity, there are intraday markets or other Energy and electricity-related products can be mechanisms that allow rescheduling units. The traded in various time frames, depending on the reasons for this rescheduling are diverse and particular market. For example: depend on market design characteristics. •• Energy, the most basic product traded on elec- •• The focus of Section 2.3 is on the markets re- tricity markets, can be procured from the very lated to the close-to-real-time actions of the long term to real time, through either physical system operator (more specifically, operations or financial contracts. carried out by the system operator after the •• The system operator may acquire ancillary ser- last market gate closure2). It is during this time vices (e.g., operating reserves that are used to frame when the detailed design of energy and ensure a continuous match between demand reserves products becomes more relevant. A and supply, even in case of contingency) in dif- proper design is crucial to provide accurate ferent time frames (including long-term, day- price signals in the very short term, and this ahead and intraday, depending on the market). will play a fundamental role in unlocking all the •• Reliability products (i.e., products that assure flexibility already available in power systems the physical availability of an adequate quan- while also providing efficient long-term signals. tity of resources in the system needed to run •• Finally, in Section 2.4, we focus on those mech- it safely without having to curtail demand) are anisms generically oriented towards directly usually procured over the long term; subse- supporting the long-term development of the quent markets may also allow the adaptation power sector. Capacity mechanisms, or any oth- of buying and selling positions. er mechanism designed to guarantee the ade- •• Renewable energy support mechanisms can quacy of the system, are introduced to achieve provide long- to short-term incentives for the the reliability target defined by the regulator. production of renewable energy. Reliability products traded in these mechanisms may be very diverse, reflecting the diver- It is not always easy to clearly separate the three sity of scarcity conditions in different systems. time frames discussed here for short-, very-short- Also, the procurement process varies (e.g., and long-term markets. All electricity markets are, centralised versus decentralised approaches) to some extent, coupled. The most relevant prod- according to the structure of the power sector. ucts that are usually traded within each time frame On the other hand, RES support schemes aim will be discussed. The analysis will start with short- at achieving a renewable penetration target. In term markets, i.e., day-ahead (DA) and intra-day this case, different designs are possible, each (ID) markets because, in most power systems, the one with its pros and cons, as analysed in detail day-ahead market serves as a reference for all the in the second part of Section 2.5. others. Later, the very- short-term (balancing or real-time markets) and the long-term market design (focusing on capacity mechanisms and RES support schemes) will be discussed.

2. The last gate closure is the moment until which market agents can modify their own programmes on the markets without the intervention of the system operator.

34 WHOLESALE MARKET DESIGN

In the remainder of this section, we introduce the Variable and small-scale generation on the sys- reasons why increasing penetration levels of re- tem dispatch requires flexibility (Holttinen et al., newables (utility scale and distributed) call for 2013; Graichen, 2015), which, among other mea- revisiting wholesale market design in liberalised sures, calls for a larger amount of dispatchable power systems for all time frames. resources capable of dealing with changing situations, whose operation is to be based on fast- 2.1.2. Why rethink the market now er and more accurate markets. Figure 2.1 illus- −− The need to rethink short-term and trates the rising need for flexibility as variable very-short-term markets today generation increases across different renewable penetration scenarios in the US Western Inter- connection. Three technologies are selected to Variable and small-scale generation and represent three different types of operating re- demand response solutions call for faster and gimes: coal plants (base load), combined cy- more accurate short-term markets. cles (mid-load) and gas turbines (peakers). An increasing penetration of RES technologies re- More than in any other kind of market, trading in sults in a higher “capacity started” among peak- power markets is largely conditioned by the laws ing units (Panel a), a decrease (even if not ho- of physics. The deployment of variable and dis- mogenous) in the number of hours that a plant tributed renewables may complicate the efficient is online each time it is started (Panel b) and an management of markets if adequate adapting increase in the number of ramps required, not measures are not put in place. Designing power only from peakers, but also from base-load units markets then, involves an ongoing search for the (Panel c) – three effects that clearly highlight the right balance between economic and engineering growing need for flexibility. efficiency, between what is desirable and what is feasible.

Figure 2.1 VRE generation and the need for flexibility

Scenarios NoRenew TEPPC HiWind HiMix HiSolar

(a) (b) (c)

300 90,000 1,000

200 60,000

500 100 30,000 Numbers of ramps Numbers Hours online per start Hours Capacity started (GW) started Capacity

0 0 0 Coal Gas CC Gas CT Coal Gas CC Gas CT Coal Gas CC Gas CT

(a) Capacity started (b) average number of hours online per start (c) total number of ramps for different plant types per year in different scenarios of RES penetration Note: The scenarios simulated are: no renewables (0% wind, 0% solar); Transmission Expansion Planning Policy Committee (TEPPC 9.4% wind, 3.6% solar); High Wind (25% wind, 8% solar); High Solar (25% solar, 8% wind) and High Mix (16.5% wind, 16.5% solar). Power plant types are coal plants (base load), gas combined cycles (mid-load) and gas combustion turbines (peakers). Results depend on the technical parameters assigned to each technology and are specific to the Western Interconnection. Source: Adapted from NREL, 2013a

35 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

The format of the bids in day-ahead markets im- of balancing products and the pricing of reserves plemented to date has traditionally been tailored (see Section 2.3). 2 to the prevailing generation mix in each particular context. The vast majority of relevant dispatch −− The need to revisit long-term decisions have been one-sided (no demand reinvestment signals sponse was expected except in very extreme situations) and could be mostly closed in the day- Long-term regulatory mechanisms (including ahead horizon. Something analogous happens capacity and RES support ones) need to be with the definition of reserves products and the designed to minimise market distortions and functioning of close-to-real-time markets. to encourage full integration of all resources. However, prevailing resources are rapidly changing. The development of renewable energy According to economic theory, perfect compet- sources (RES) has fostered the proliferation of itive markets lead to the most efficient results, distributed generation, giving rise to a signifi- both in the short and in the long term. In real cas- cant number of new small-scale agents and busi- es, however, the presence of market failures im- nesses that need proper accommodation in the plies that this ideal outcome is often not reached. market. In addition to this and in parallel, enabling information and communication technolo- When there is no way to fix these failures to al- gies are increasingly allowing demand to receive low the market (left to its own devices) to achieve more precise market signals and to react to them, the welfare-maximising outcome, some addition- which will empower consumers to play a major al regulatory mechanisms can help the market role in the energy transition. In this new context, meet this goal. These mechanisms may look to there is growing consensus around the fact that reinforce the short- and/or long-term signals that short-term and very-short-term3 markets need to agents perceive in order to improve system secu- be improved and adjusted to meet future needs. rity of supply or may seek to incentivise certain technologies (e.g., wind or solar). The major refinements needed depend to a large extent on the particular type of market design Capacity remuneration mechanisms in place. To simplify, there are two main ways of No consensus has been reached yet regarding conceiving short-term electricity markets (Bat- instruments for the remuneration of capacity. A lle, 2013): the US Independent System Operator good number of markets, mainly on the the Amer- (ISO) approach and the Target Model for the Eu- icas (Batlle et al., 2015) opted to implement a di- ropean electricity market.4 verse array of explicit capacity mechanisms from the outset, while in other regions (Europe is the The short-term market design elements consid- paradigmatic example), this was not initially the ered pivotal to the current transition include the case.5 However, several countries in Europe have time frame of markets, bidding formats, clearing now implemented, or are in the process of imple- and pricing rules and the integration of energy menting capacity mechanisms (ACER, 2013; AC- and reserve markets (see Section 2.2). ER-CEER, 2016a), as also evidenced in Figure 2.2.6 In very-short-term markets, the design elements in need of revision are the definition of balancing responsibility (for conventional and RES generation), the imbalance settlement, the definition

3 Balancing markets in the European Union and real-time markets in the United States. 4 As defined by CACM Guideline (Commission Regulation 2015/1222) and draft Electricity Balancing Guideline. 5 Although all sorts of implicit regulatory safeguards have been applied indirectly in order to guarantee the security of supply in these energy-only markets. 6 The European Commission has highlighted the need to use a harmonised adequacy assessment methodology, in order to avoid the design of CMs at country level to be a potential threat to to the development of the internal energy market. For further details, see the Winter Package released on 30 November 2016 (https://ec.europa.eu/energy/en/news/commission-proposes-new-rules-consumer-centred-clean-energy-transition).

36 WHOLESALE MARKET DESIGN

The deployment of RES has been argued by some ditional plants. Indeed, in a competitive electrici- as the new and key factor to justify the need for ty market – given the standardised nature of the the implementation of capacity mechanisms. The product – the market price is equal to the margin- argument more commonly raised is that high al cost of serving an incremental unit of demand. volumes of variable renewables in markets with Therefore, as long as there is enough capacity to overcapacity are driving down short-term prices serve the demand, market price cannot exceed and load factors of conventional generation. This the marginal variable cost of the system. Because makes it more difficult to forecast the frequency electricity cannot be stored economically, only of scarcity events that, in a competitive “energy when capacity is scarce with respect to demand only” market, are the ones that allow for the re- can prices surge over variable cost, contributing covery of the fixed costs of non-incentivised tra- to the recovery of fixed costs.7

Figure 2.2 European power systems with (or in the process of implementing) capacity mechanisms

No CM (energy only market) CM proposed/under consideration CM operational

Strategic reserve (since 2007)

Strategic reserve Strategic reserve (since 2004)-gradual (since 1 november 2014) phase-out postponed 2025

Capacity auction (since 2014-ﬁrst delivery Strategic reserves (Envisaged in 2017) in 2018/19)

Strategic reserve Capacity payments (from 2016 on, for 2 years, (since 2007) considering with possible extension reliability options for 2 more years)

Capacity requirements Strategic reserves (certiﬁcation started 1 April 2015) Reliability options (the date for the ﬁrst auction Capacity payments has not been set. First (since 2008) delivery of contracted Tendering for capacity capacity is expected in 2020) considered but no plans

Tender Capacity payments (since November 2013) (since 2010 partially suspended between May 2011 and December 2014) New Capacity Mechanism under assesment by DG COMP (Capacity payments from 2006 to 2014)

Note: In Germany, there are three (envisaged) schemes: Climate Reserve, Network Reserve and a Strategic Capacity Reserve. The first is not considered to be a CM; the second could be, and the third is a CM. The Strategic Capacity Reserve is envisaged to be implemented in 2017, if the necessity is demonstrated. The envisaged CM in Poland for after 2016 includes generation units tendered by the TSO, which would definitely have been decommissioned by the end of 2015. This scheme has the characteristic of a Strategic Reserve CM. The term “country” as used in this material also refers, as appropriate, to territories or areas. Source: Adapted from ACER-CEER, 2016a, based on NRAs (2016) and European Commission’s report on the sector inquiry into CMs (2016).

7. “In an energy-only market, the signal for investment relies on high prices that materialize in moments of excess demand (these are called scarcity prices and moments of excess demand are scarcity scenarios): whenever there is a scarcity scenario, prices are allowed to rise so that generators start earning ‘scarcity rents’ that are high enough to cover their fixed costs of capital and induce new investment/new entry in the market. […]” see European Commission, 2015a.

37 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

In this context, many stakeholders have ques- for generation adequacy and renewable devel- tioned whether short-term market prices alone, opment. The goals are to send proper long-term 2 even if regulations allow market prices to reach signals, to avoid interfering and distorting market very high levels in scarcity events, are a reliable functioning as much as possible, and to encour- enough signal to encourage adequate capacity in- age the integration of all resources. vestments. In fact, the need to reinforce long-term signals does not arise only from RES deployment 2.2. SHORT-TERM MARKETS itself, but also from the regulatory uncertainty (e.g., around the kind of support to be provided in the future) around deployment and the effects of Short-term energy market design needs to be this uncertainty on wholesale markets. enhanced and refined at all levels, particularly timelines, locational granularity of prices, RES support mechanisms bidding formats, clearing and pricing rules For many years, there has been a consensus on and integration with reserves. the need to implement long-term RES support mechanisms because of the market’s inability to internalise a number of externalities. However, Short-term auctions are the core of electricity RES support mechanisms, and therefore RES in- wholesale markets. At these auctions, bids by vestments, have not always benefited from stable generators and consumers are matched with the remuneration frameworks. This lack of regulato- textbook objective of determining not just who ry predictability, among other factors, affects the sells and who buys, but also market clearing pric- decisions of potential RES investors (and also the es (for each time interval in the auction timeline decisions of investors in conventional generation scope and for both market sides: generation and technologies). demand) – that is, as described in figure 2.3, the price at which demand equals supply, and no A discussion on whether the need for such mech- more demand is willing to purchase and/or sup- anisms in each particular context is justified or ply willing to sell. These short-term electricity not falls outside the scope of this study. Here, the prices are essential since they represent the ref- focus is on providing recommendations about erence for the longer-term markets (which help how to design these support mechanisms, both drive the system expansion).

Figure 2.3 Market clearing price and market clearing volume setting in electricity auctions

$/MWh

DEMAND SUPPLY

Market clearing price

Market MWh clearing volume

38 WHOLESALE MARKET DESIGN

In electricity markets, trades concern an underly- erator. But the two approaches are increasingly ing commodity which is subject to many complex converging and promise to continue doing so. physical constraints (operation, network, security, etc.). Different approaches to dealing with these −− Short-term market design elements to be re- physical complexities have resulted in (some- considered times significantly) different market designs. We have pointed out that the recent proliferation of resources with new characteristics has raised In practice, as briefly analysed in Box 2.1, there concerns regarding the need to improve the cur- are two main approaches to accounting for all rent market design. Among the different design these physical constraints: the US and the Euro- features of short-term markets, we focus here pean Union (EU) models. The major difference on those that we believe deserve more attention between the two is the degree of separation be- from stakeholders and policy makers: tween the market operator and the system op-

Box 2.1 The major difference between US and EU market designs

The major difference between US and EU market designs requirements set by the ISO (e.g., different types of re- is the degree of separation between the roles of market serves) can be co-optimised and priced along with en- operator and system operator. In certain jurisdictions, the ergy. role of the system operator is less pervasive than in oth- At the other extreme, the model currently used in most ers, under the argument that the system operator func- −− European member states, which we can call the EU pow- tion should be limited to ensure reliability in an attempt er exchange (PX) approach, originally aimed at a simpler to maximise the range of the market. The debate on allo- consideration of the physical reality along with a major cation of responsibilities between the market and the sys- decoupling of the system operator’s responsibilities and tem operator has been central since the outset of market the spot market functioning. Historically, European Pow- restructuring (see, for instance, Hogan, 1995). er Exchanges were designed as financial platforms for The predominant models implemented in US markets and agents to buy/sell energy with the supposed objective of in force in the majority of EU member states – with some maximising liquidity, in an attempt to isolate (as much as exceptions, including Ireland until very recently (CER, possible) this trading from the complexities of the net- 2014) and Poland (Siewierski, 2015) – represent two dif- work. In line with this separation between the market and ferent views on whether or not to carry out this separa- the system operators, the European approach tends to tion. allow generators to bid on a portfolio basis (aggregated bids from different facilities of the same company).8 −− In the United States, the integration of physical constraints in the clearing process and the involvement of −− In a nutshell, as will be discussed throughout the chap- the independent system operator (ISO) in the markets is ter, the influence of the system operator to determine the probably as significant as is possible. The ISO oversees real-time scheduling throughout the entire process is, in both activities, market and system operation, not only practice, reflected in a number of basic design elements, from an institutional perspective, but also from the op- for example, the point in time at which the last gate clo- erative standpoint, as energy trades are jointly cleared sure is set (i.e., the moment until which market agents can with the electricity system security procedures. This is modify their own programmes without the intervention considered necessary, not only to reliably operate the of the system operator), the integration of the physical system, but also to guide market agents towards the op- constraints in the clearing process, the amount and di- timal dispatch (from the system operator’s perspective, versity of reserves defined by the system operator and taking into account technical and reliability constraints). the mechanisms to acquire them, or even the role of the This model revolves around the concept of bid-based, se- system operator as aggregator and balancing party re- curity-constrained economic dispatch (SCED), where the sponsible for RES.

8. Despite being the major trend in Europe, portfolio bidding is not allowed in some European systems, including, for instance, in Spain and Portugal.

39 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

•• The time frame of markets, both for energy and 2.2.1. Time frames of markets, reserves, which is fundamental to allow market dispatches and prices 2 mechanisms to respond quickly to fast-changing conditions. As variability increases, the time specificity of • The locational granularity of prices and sched- • produc ts traded in the markets needs to be ules. increased. This leads to the need for better- • The so-called bidding formats. i.e., the way • adapted and more flexible market timelines. generators are allowed to submit their offers in the market (not only in the day-ahead but also closer to the real-time markets). One central element of the short-term market de- •• How markets are cleared and settled, i.e., how sign that needs to be revisited to better accom- it is determined who produces and who con- modate RES is the time frame of the energy and sumes and at what price. reserve markets. The time frame of the system •• How to define and procure operational reserves, operator’s actions (i.e., reliability-related pro- and particularly how to guarantee that the mar- cesses involved in the delivery of electricity) also ket reflects the interdependence between the needs reconsideration. energy and reserve products, especially under The design of market time frames is mainly in- conditions of scarcity. fluenced by the physical and technological char- Some of these design features are relatively new acteristics of the power system. Since the outset (e.g., the need to adjust the market timeline to of the electricity market liberalisation, there has fit the new flexibility requirements or the need to been a general consensus (with some exceptions) define new bidding formats), while other refine- that, in thermal-dominated systems, it is conve- ments (for example, the need to revisit clearing nient to implement a day-ahead energy market and pricing rules) target deep-rooted efficiency and also day-ahead reserve markets. Energy and issues that have been acknowledged for years, reserves can be procured: 1) within the same auc- but whose effect has been exacerbated by the tion mechanisms or 2) by means of different and new market conditions. sequential mechanisms. Although most of the previous design elements For a short-term energy market, one day is suffi- of short-term markets are related to some ex- cient to both forecast demand consumption with tent, for the sake of clarity, we will try to deal with reasonable certainty and to schedule thermal them one at a time and separately. plants’ commitments. Meanwhile, for a hydropower-dominated (energy-constrained) system, the economic dispatch and the resulting prices can be calculated over larger horizons (e.g., on a weekly basis, as done in Brazil). This is evidence of the significant role that electricity systems’ technological characteristics play in deciding optimal time frames.

40 WHOLESALE MARKET DESIGN

From the day-ahead market to real time, addi- market, where resources are purchased or sold tional markets and mechanisms are needed to al- by the system operator (which is the counterpart low agents to adapt their programmes to chang- of all transactions). As described later, in the Eu- ing conditions. Increasing shares of wind and ropean model, there are intraday markets where solar generation, in particular, result in increas- energy products are traded directly between ing volumes of intraday trading necessities, and market players. the need to adjust production schedules to the The first difference between both schemes in most recently updated forecasts. This requires the short term lies in the way operating reserves the market time frame to adapt to fully exploit (generation capacity readily available to the sys- the potential of variable renewable resources. tem operator for solving contingencies) are pro- Figure 2.4 offers an overview of market time cured: frames in the US and European contexts. This •• In the case of ISO markets, reserves are usually representation is simplified for clarity and main- procured simultaneously with the energy mar- ly focuses on the most relevant differences be- ket (in the US ISO’s terminology, they are often tween the two paradigms. co-optimised). The figure indicates when each process takes •• In Europe, operating reserves are organised by place and the time scope covered. For example, the system operator after9 the energy market the day-ahead market in Europe takes place in results are known. Therefore, the procurement the morning of the day prior to electricity deliv- is sequential in this case. (The drivers and im- ery (operating day – 1) and covers all 24 hours of plications of this alternative are discussed in the delivery day (operating day). Figure 2.4 also Section 2.2.5.) specifies whether the process is an energy market, where products are traded directly between market players, or a system operator’s centralised

Figure 2.4 Short-term market time frames: An overview

Operating Day -1 Operating Day

Day Ahead Energy Reliability Unit (& Reserve) Commitment Trading/Operating horizon Market (ISO dispatch correction)

ISO model ISO System Operator’s Market (Real-Time Market [RTM])

Day Ahead Day Ahead Ancillary services (Reserve Energy Trading/Operating horizon Market procurement, Congestion management...)

Intraday Energy Markets

European model European System Operator’s Market (Ancillary and Balancing Markets)

0:00 Day D Energy Markets: traded directly between market players System Operator’s centralised markets Trading/Operating horizon

9. Reserve resources can also be procured prior to the day-ahead energy market through longer-term agreements (e.g., through monthly tenders/auctions).

41 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

After the day-ahead market and the reserve pro- correct their deviations without any type of in- curement process(es), the system operator takes tervention from the system operator. After that 2 actions to ensure the reliability of the system. The point in time, the final binding production sched- timing and scope of these actions are similar in ule is determined for all participants, and only the both contexts.10 These actions are needed after system operator can adjust any deviation. The the day-ahead market because this market does timing of the last gate closure represents the di- not account for all the physical complexities of viding line between markets and pure system op- the power system. As mentioned above, in the EU erations. case, this lack of detail is partly the consequence •• In Europe, the last opportunity for a market of decoupling the system operator’s responsi- agent to change its schedule in hour h (or in a bilities from the functioning of the spot market, market time period x), without the system op- isolating as much as possible the energy trading erator’s intermediation, is the closure of the last from the complexities of the network-constrained intraday market session in which energy can be dispatch model. This separation comes at a cost, traded for that delivery period h (see following which is the appearance of potential infeasibilities section on intraday settlements), see Box 2.2. that have to be later solved by the system opera- •• In the United States, this dividing line between tor. In the United States, although the objective is markets and purely system operations is slight- to include all the details in the energy market pro- ly different. Once the day-ahead market closes, cesses, it is still found that today, “the full alter- the ISO calculates the so-called reliability unit nating current representation of the transmission commitment (RUC), which corrects the sched- system cannot currently be explicitly included ule for the day after. The ISO first clears the in the market software despite the use of state- day-ahead market and develops the necessary of-the-art computational tools” (FERC, 2014c). corrections. It checks either for eventual trans- Some operational and reliability considerations mission constraints that could not have been remain impossible to incorporate into the day- fully captured by the security-constrained eco- ahead market process. As a result, the system nomic dispatch optimisation model, or seeks operator also needs to undertake some follow-up to adjust cleared bids to the ISO’s forecast of actions in US electricity markets. load and RES production or as a result of the Apart from the previous issues, which are analysed co-optimisation of both the energy and reserve in subsequent sections, there are three fundamen- markets (see Section 2.2.5). In principle, mar- tal design decisions related to the time frame of ket agents are not supposed to re-adapt their markets: 1) the definition of the time threshold bids from that point on, so to some extent this within which bids have to be presented in intr- would be the gate closure. However, market aday energy markets and beyond that only the agents, under certain specific and well-justified system operator can take action/dispatch re- circumstances, are allowed to update their day- sources through its centralised (balancing) mar- ahead offers. This reoffering period may end a ket (the so-called last market gate closure or sim- few hours to a few minutes before real time. ply the gate closure); 2) the timeline and format The increased uncertainty on real-time opera- of intraday energy markets (run by the market tions, partially as a result of the growth of vari- operator in the European Union and by the ISO in able renewable energy, is stimulating this de- the United States) and 3) the settlement period. bate. And again, the discussion revolves around These designs are discussed next. the boundaries between the market and the sys- Reoffering period and gate closure tem operator’s competencies. As discussed in −− the next subsection, in principle, widening the Until the so-called last gate closure, market scope of the market provides market agents with agents are allowed to balance their positions and

10. Although some ISOs integrate several reliability processes into the day-ahead market, as discussed in Section 2.2.5.

42 WHOLESALE MARKET DESIGN

incentives to do their best to minimise their im- ity. Therefore, having the gate closure away from balances. This can be of utmost relevance as RES real time would be preferable to faciliate the task penetration rises since allowing market agents of the system operator and to reduce system se- to update their offers closer to real time would curity costs. In deciding the point in time when improve the accuracy of their forecasts. (Miller, market participants have the last opportunity to 2015)11 On the other hand, forcing the transmis- change their schedule, benefits (from a RES pro- sion system operator (TSO) to operate too close ducer point of view) and costs (from the system to real time could require procuring reserve re- operation point of view) should be compared. sources in larger amounts and/or of higher qual-

Box 2.2 Short-term electricity market timeline in Italy

Day-ahead market tion are made known by 12.55 p.m. on the day before the day of delivery. The MGP’s auction selects bids and offers In the day-ahead market, electricity is traded one day be- based on economic merit-order criterion, taking into ac- fore the operating day. In a wholesale market, electricity count transmission capacity limits between zones. can be traded day-ahead bilaterally (over-the-counter trading) or on the day-ahead power exchange. Intra-day market

In Italy, the day-ahead-market, called the MGP, is an auc- The intra-day market (MI) allows market participants to tion market, not a continuous-trading market. Bids and modify the schedules defined in the MGP by submitting offers can be submitted to the MGP beginning at 8:00 additional supply offers or demand bids. The MI takes a.m. on the ninth day before the day of delivery until 12:00 place in seven sessions: MI1, MI2, MI3, MI4, MI5, MI6 and p.m. on the day before the day of delivery (Day — 1), when MI7. The schedule of each MI is as shown in the Table 2.1. the auction process begins. The results of the MGP aucº-

Table 2.1 Bid sessions for intraday market in Italy

MI session Opening time Closing Time Result

MI 1 12.55 pm (Day -1) 3.00 pm (Day -1) 3.30 pm (Day -1)

MI 2 12.55 pm (Day -1) 4.30 pm (Day -1) 5 pm (Day -1)

MI 3 5.30 pm (Day -1) 11.45 pm (Day -1) 00.15 am

MI 4 5.30 pm (Day -1) 3.45 am 4.15 am

MI 5 5.30 pm (Day -1) 7.45 am 8.15 am

MI 6 5.30 pm (Day -1) 11.15 am 11.45 am

MI 7 5.30 pm (Day -1) 3.45 pm 4.15 pm

Source: Gestore mercati energetici, 2017.

11. But this may be relevant for thermal plants too since it would allow them to update thermal generation offers so they can represent intraday changes in gas market prices.

43 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

−− Intraday settlements (see also Box 2.4). Intraday markets produce pric- Intraday markets in Europe perform two relevant es used to settle the incremental changes execut- 2 functions. First, as shown in Box 2.3, they allow ed in each market. generators to make incremental adjustments to In the United States, ISOs perform intraday ad- their energy schedules (resulting from the day- justments according to updated forecasts on the ahead market), which improves the efficiency of evolution of the system (mainly related to load the initial day-ahead programme. This can miti- forecasts and generation forced outages). These gate the potential inefficiency of the day-ahead actions are adapting to a larger penetration of redispatch due to the limitations of the European newable energy sources. These processes are not bidding formats (see Section 2.2.3). Second, they equivalent to European intraday markets in one recover the balance between demand and sup- key aspect: generally, no intraday price signals ply if the forecasted system conditions change. are calculated in the United States associated with For example, intraday markets are fundamental to these actions. This way, the US intraday actions do accommodate renewable energy forecast errors, not establish binding economic transactions (for and most European intraday markets have recent- they do not produce binding prices). All deviations ly been modified to improve their performance

Box 2.3 Intraday markets and schedule adjustments

Intraday markets are useful platforms to accommodate markets and which, also due to the impossibility of sub- possible forecast errors (e.g., on demand or on renew- mitting multi-bid offers, could be technically unfeasible. able energy production) as well as to modify the produc- The need for intraday markets is closely linked to the fact tion’programs of thermoelectric power plants (as shown that deviations from programs (imbalances) are penal- in Figure 2.5), defined as a result of day-ahead energy ised. The intraday market therefore helps reduce the risk of imbalances.

Figure 2.5: Example of schedule adjustments in intraday market

Unit X production program, Unit Y production program, as deﬁned in day-ahead market (MWh) as deﬁned in day-ahead market (MWh)

20,000 30,000 25,000 15,000 20,000 10,000 15,000 10,000 5,000 5,000 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Unit X production program, as changed in Unit Y production program, intraday market (MWh) as changed in intraday market (MWh)

20,000 30,000 25,000 15,000 20,000 10,000 15,000 10,000 5,000 5,000 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

44 WHOLESALE MARKET DESIGN

from the day-ahead programme are settled later tlement system in ISO markets dilutes these sig- at the real-time price,12 i.e. the price that represents nals since all deviations are settled at the real-time the value of the electricity in real time (taking into price. The important role of intraday markets and account the actual condition of the system), cal- signals has also been confirmed by other analy- culated every five minutes through a unit-commit- ses on market design. For example, the Interna- ment-like model (similar to the one used for the tional Energy Agency (IEA, 2016) pointed out that day ahead but with updated information). “transparent intraday prices are necessary to in- form all market participants about the cost of serv- Ideally, generators deviating from the day-ahead ing the next megawatt”, and also that “the design programme should receive signals representing in North America could be further improved by the cost associated with the deviations. These providing greater transparency of the evolution of costs can be lower if deviations are known in ad- locational prices during the intraday timeframe”. vance. For example, a renewable energy source having a lower output than anticipated in the −− Continuous versus discrete intraday markets day-ahead forecast could require other resources in Europe to fill the gap. If the deviation is known only a few European intraday markets, pending a unified de- minutes before real time, this deviation will likely sign (XBID project), have been implemented by require fast-start units or expensive flexible gen- Member States using two methods: 1) discrete erators, but if notified several hours in advance, auctions, in which auctions are called at specific the operator has enough time to adjust the out- predefined times (Figure 2.6, upper part) and 2) put of inflexible resources more economically. continuous trading, in which bids can be submit- Intraday prices in Europe capture the different ted and matched at any time before gate closure value in times of deviations, while the two-set- (Figure 2.6, lower part).

Figure 2.6 Time frame of European intraday markets

Operating Day -1 Operating Day

Day Ahead Energy Market Day Ahead Intraday Energy Market (IM) Market IM1 IM2 IM3 IM4 IM5

Intraday Auctions Intraday IM6 IM7 TSO’s Ancillary and Balancing Markets Day Ahead Energy Market Day Ahead Intraday Energy Market (IM) Market

IM1

IM2 EU Intraday Markets: Auction vs Continuos trading Continuos vs Auction Markets: EU Intraday Intraday Continuos Trading Continuos Intraday

TSO’s Ancillary and Balancing Markets

0:00 Day D Energy Market Bidding Energy Market Bid Selection Intraday Energy Market Bidding Period period and Bid Selection Trading horizon Energy Market Process TSO’s Market Process Input for TSO’s market

12. The day-ahead and the real-time markets make up the so-called two-settlement system. 45 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

As shown in figure 2.6, auctions provide for dif- Flexibility versus liquidity ferent times for bid submission and selection, Calling auctions at specific, predefined times 2 whereas with continuous trading, bids can always provides market agents with limited flexibility to be submitted, and once submitted, they can be change programmes since the market can only selected at any time. As a consequence, intraday capture these changes at the subsequent auc- auctions have different sections open at the same tion. From a flexibility point of view, this is less of time, each collecting different bids and with its an issue if auctions are held frequently. own results, even though the trading horizon can partially overlap. Intraday continuous trading, in- The potential advantage of discrete auctions is stead, is characterised as having just one session that they may provide higher liquidity since they at a time, with a shortening of the trading horizon concentrate transactions reflecting all the accu- (i.e., of the hourly products that can be traded) mulated events since the last session. Clearly, as time passes. increasing the frequency of these auctions may negatively affect their liquidity. Discrete auctions, Figure 2.7 presents the intraday market designs therefore, need to be configured depending on implemented across different European pow- the system characteristics, and they need to ener systems, which are based on two criteria: 1) sure sufficient liquidity. whether these markets are continuous or auction-based and 2) which entity owns the trad- On the other hand, continuous trading provides ing platform used to carry out those transactions greater flexibility (since trading is possible at any (i.e., is the ownership based on a power exchange time). However, as is the case with too frequent [PX], TSO or bilateral?). discrete auctions, continuous trading may result in insufficient liquidity. Each of the two above-mentioned alternatives (i.e., continuous and auctions) has the following characteristics, advantages and disadvantages.

Figure 2.7 Intraday trading arrangements across Europe

Spain Poland: Italy physical Auction Germany: 15 min auction Poland: ﬁnancial

Germany Continuous The Kingdom of Netherlands

Bilateral Exchange - based TSO - based redispatch

Source: Adapted from Neuhoff et al, 2015

46 WHOLESALE MARKET DESIGN

Pricing cross-border capacity es in the level of consumption/production from in continuous markets RES are managed through regulation resources). Debate continues as to whether intraday auc- However, the product traded in energy market is tions can more efficiently price the cross-border standardised in the sense that it is assumed that, transmission capacity compared to continuous in order to comply with the commitment to sup- trading. The major problem is that in a continu- ply/withdraw the energy sold/purchased in the ous trading context, cross-border transmission market – i.e., in order to be balanced – it is suf- capacity is allocated on a first-come, first-served ficient to do so “on average” within predefined basis; this does not allow scarcity to be properly temporal windows, named settlement periods priced in transmission. In other words, transmis- (also referred to as trading periods) that are typ- sion capacity is allocated little by little at a zero ically one hour long. It is the TSO’s role to assure price until it becomes congested (and at that pre- that the system is continuously balanced; to do cise moment, there is no more available spare ca- so, all balancing resources have to divide up in pacity to allocate). sub-periods (e.g. 15 or five minutes) their previously assumed production commitments. Then A potential solution: hybrid design the TSO in the balancing and ancillary services A hybrid design, combining a flexible continuous markets negotiates changes in their net produc- market with a number of discrete auctions with tion level in each of those sub-periods (or dis- concentrated liquidity, may represent a suitable patch periods), and this updated program will trade-off. The question is: to what extent can determine the final schedule based on which un- adding auctions to continuous intraday trading balancing penalties are applied. actually improve the performance of the market? Given that for balancing resources (typically ther- In Germany, the implementation of an addition- mal power plants) it is more efficient to produce al local intraday auction (called at 3:00 p.m.) to at a constant level, they will try to uniformly dis- complement continuous trading is assessed by tribute the planned output of a settlement peri- Neuhoff et al. (2016). The analysis shows how od through the different sub-periods. This means trading volumes, liquidity and market depth13 in- that the higher the demand (or the RES produc- creased after the auction’s implementation. This tion) volatility within a settlement period, the is in line with previous observations showing more actions the TSO will have to implement to that liquidity in European power markets based balance the system. Thus, as the RES production on continuous intraday trading (e.g., in the cen- quota increases, it becomes more important to tral western states) is lower than in markets hav- reduce the size of the settlement period for VRE ing auction-based intraday trading (e.g., Italy and plants as well. Spain). The length of the settlement period is particularly These discrete auctions can be called either at relevant for trades close to real time. This is why specific predefined times or upon the occurrence we first discuss the role of the settlement period of particular events (e.g., a forced outage of a in the very-short-term markets and then extend large plant, a high forecast error, etc.). the discussion to the day-ahead market.

Settlement and dispatch period Settlement period in intraday −− and very-short-term markets A safe and reliable system requires that energy supply always equals energy demand. As the In the United States, the real time market produc- demand, as well as the production from renew- es five-minute dispatch instructions (Figure 2.8), able sources, continuously fluctuates over time, and prices are also computed every five minutes. the (marginal) cost of producing electricity varies However, all ISOs, except the New York Indepen- almost continuously (even though small chang- dent System Operator (NYISO), have traditional-

13. Market depth is the market’s capability to sustain large market orders without affecting the market price. It can be measured as the amount of electricity that is offered at certain prices during certain time periods.

47 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 2.8 The time frame of US real-time markets

2 Operating Day - 1 Operating Day

Day Ahead Reliability Unit Energy Commitment (& Reserve) Trading/Operating horizon (ISO dispatch correction) Market

System Operator’s Market (Real-Time Market [RTM]) ISO model ISO

0:00 Day D

RTM RTM RTM 5 minutes RTM RTM RTM

Energy Markets: traded directly between market players System Operator’s centralized markets Trading/Operating horizon

ly calculated average hourly prices to settle re- of reserve needs and use – long periods create al-time transactions to simplify the metering and systematic deviations as a consequence of artifi- settlement process. This simplification is widely cially breaking a continuous and smooth demand viewed as relatively inefficient, and the Federal into discrete steps. Figure 2.9 shows the system Energy Regulatory Commission (FERC) recently imbalance in Germany for every minute of the issued Order 825 (FERC, 2016) requiring all ISOs day during 2011. Imbalance deviations are larger to settle energy transactions in the real-time mar- around the end and the beginning of each hourly ket in five-minute intervals. period. These leaps around full hours can also be observed in grid frequency (see Weißbach and Recent concerns regarding the integration of re- Welfonder, 2009). As previously mentioned, this newable generation (due to short-term variabil- is due to the fact that, conventionally, the pro- ity, as discussed earlier) had already motivated duction/consumption programs are uniformly sub-hourly settlements in some US ISOs – such distributed during the settlement (or dispatch) as in the California Independent System Oper- period, whereas the actual demand and produc- ator (CAISO), where 15-minute settlements are tion changes continuously within each settlement already implemented; or in the Midcontinent In- period. dependent System Operator (MISO) and the Independent System Operator New England As a consequence of this inefficient use of re- (ISO-NE), where currently changes are being im- serves, Germany opted to reduce the settlement plemented. However, taking into consideration period in intraday trading down to 15 minutes, i.e. that “a movement to sub-hourly settlements may at time when VRE production and demand can be costly and difficult to accomplish in a short pe- be better predicted (see box 2.4), though some riod of time” (PJM, 2015b), the FERC allows addi- EU systems still use longer settlement periods in tional time for full implementation of this reform. intraday markets. The intraday market is usually aligned with the balancing market, where TSOs Dispatches and prices are less granular in Europe- charge for imbalances between supply and de- an systems. Using long settlement periods (e.g., mand, measured in 15-minute intervals (Germany, one hour) has relevant consequences in terms

48 WHOLESALE MARKET DESIGN

Figure 2.9 Average German system imbalance for every minute of the day during the year 2011

1,000

500

-500 (Mean) [MW]

-1,000

-1,500

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Time (Hours of the day)

Source: Adapted from Hirth and Ziegenhagen, 2015

the Kingdom of Netherlands, and Italy for quali- its Day-Ahead market could significantly increase fied units), 30-minute intervals (France, the Unit- the computational time” (NYISO, 2015). Other ed Kingdom) and hourly intervals Italy, Poland and ISOs do not see clear advantages: “ISO-NE be- Spain for non-qualified units) (Neuhoff et al., 2015). lieves that the potential benefits of sub-hourly settlements in the day-ahead market are differ- The European Commission (2016a) has recently ent, and likely to be much smaller than in the re- proposed a regulation explicitly requiring market al-time market.” The potential drawbacks include operators to allow trading energy in intervals as those noted by MISO (2015): “Due to the imper- short as the imbalance settlement periods (i.e., fect knowledge of real-time net load or unex- aligning intraday and balancing markets), which pected outages one day ahead, scheduling at the is at 15-minute intervals in all Member States. sub-hourly level may actually worsen the consis- While this proposal will improve granularity in tency and cost-efficiency than scheduling at the electricity markets, its projected implementation hourly level with some robustness”. date – January 2025 – is not very ambitious. In the European context, the above-mentioned Settlement period in the day-ahead market European Commission (2016a) proposal requires A more debated topic is whether subhourly set- the use of 15-minute intervals in both the day- tlements should be used in day-ahead markets. ahead and intraday markets. This measure will CAISO (2015) sees many advantages in doing have the positive effects of increasing granular- so: “Among the possible benefits of sub-hourly ity in wholesale markets and aligning electric- settlements in the day-ahead market are align- ity products from the day-ahead market with ing unit commitment decisions between the day- the balancing market (therefore preventing in- ahead and real-time timeframes, […] improve efficient arbitrage opportunities). The poten- forward scheduling of variable energy resources tial drawbacks of sub-hourly settlements found and help align the CAISO ramping requirements in ISO markets (mainly, excessive computational with resource operating characteristics”. Howev- complexity) should also be expected in Europe, er, a more granular day-ahead market could re- although the use of 15-minute intervals (instead quire excessive computational complexity; for of five minutes) may alleviate the challenge. example, “The NYISO has concerns that implementing sub-hourly commitment scheduling in

49 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

2.2.2. Locational granularity of prices Box 2.4 Intraday markets and forecast accuracy and schedules 2 Market participants submit their offers in the day ahead market based on their assumption of the ac- High shares of VRE deployment, particularly tual system conditions and on their forecasts for re- wind, might lead to an increasingly constrained newable energies. As the time of delivery approach- transmission system. It is reasonable to assume es, the accuracy of renewable generation forecasts that zonal pricing mechanisms may lead to increases (see figure 2.10 upper part). The intraday market enables market participants to better reflect higher inefficiencies with high penetration these forecasts in their positions. levels of VRE, both in the short and in the long run, compared to nodal pricing. Moreover, quarter-hourly products are traded in the intraday market which enable a better approximation of the real demand ramps and variable A number of regulatory options address the is- renewable energy than the hourly products at the sue regarding the allocation of limited transmis- day-ahead market. This is especially important since sion capacity for transactions among players un- imbalance settlement periods are on a quarter-hour- der normal market conditions.14 These options fall ly basis. Figure 2.10 (bottom part) shows the differ- into two main groups: pricing algorithms that in- ence between hourly and quarter-hourly products volve a detailed representation of the transmis- and their accuracy in representing actual feed-ins. sion network, and those that consider a simplified one (this is true for all market segments, even if Figure 2.10 Approximation of realized load different approaches can be used for each seg- and forecast errors in intraday markets ment):

•• Nodal pricing applies security-constrained eco- Day-ahead Intraday forecast error forecast error nomic dispatch to derive locational marginal prices – i.e., the prices paid for the energy io n consumed or generated at a given transmission

ealiza t node (a substation or any other element that creates a discontinuity in the grid). Nodal ener- eca st / r r

o gy pricing provides an accurate description of F the technical and economic effects of the grid Time on the cost of electricity. It implicitly includes

Day-ahead forecast Intraday forecast Actual realization the effect of grid losses and transmission congestion, internalising both effects in a single Hourly products 15-min products value (monetary unit per kilowatt hour [kWh]) that is different at each system node. Therefore, nodal prices are perfectly efficient signals for economic decisions concerning the short-term operation of generation and demand since they Feed-in / load correctly convey the economic impact of losses and constraints at all producer and consumer Time locations. • With zonal pricing, the power system is ad- Source: Adapted from TenneT, 2017 • ministratively divided into zones, within which little congestion is expected. Price differentials reflect only transmission congestions considered relevant by the regulator or the system operator. Different degrees of spatial resolution are possible (France and Germany feature only one price zone, while Italy has ten). Accurately

14. See, for example, Rivier et al. (2013).

50 WHOLESALE MARKET DESIGN

defining stable zones is not a straightforward must withdraw from the system and which are task. Zones do not match national borders and to be included. Energy removed to solve a net- can change hour by hour alongside the chang- work constraint may be paid at the respective ing supply/demand equilibrium. An example is agent’s bid price (if a specific bid related to the provided in Figure 2.11, where the effect of wind constraint-solving mechanism is in place), at the output on zone definition is shown. opportunity price (the energy market price less the price of the agent’s bid in the energy market) With zonal pricing (as with single-node ap- or not at all. When additional energy is request- proaches that consider an entire grid as a single ed, it is normally paid at the respective agent’s node), a re-dispatch may be necessary to solve bid price. congestions (this is true for all market segments). In the (supposedly) few cases in which grid con- The pros and cons of nodal and zonal pricing have straints are detected, the system operator re-dis- long been debated by experts and policy mak- patches the system, determining which players ers. Several issues are to be taken into account

Figure 2.11 Simulation of suitable zones for zonal pricing in Europe (maximum wind)

500 N

400 N

W 300 E

00 200 E 100 E

Maximum wind No wind

10 28 46 64 82 100 Energy price at nodal level (Euro/MWh)

Source: Adapted from Neuhoff, 2015

51 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

when estimating the benefits of these two ap- RES and the efficiency of locational granularity proaches: the efficiency of the resulting price sig- The planning process in liberalised electricity sys- 2 nals (both in the short and long term), the com- tems may be more difficult than within vertically putational burden and implementation costs, the integrated utilities because decisions about gen- hedging complexity and the impact on the liquid- eration expansion are the result of market forces ity of long-term markets and geographical con- influenced by policies and regulations rather than sumer discrimination. In usually congested elec- centralised planning. tricity networks, the benefits from nodal pricing seem to outweigh the disadvantages (Neuhoff When generator build times are shorter than and Boyd, 2011). On the other hand, in densely those for transmission, planners are forced to ei- meshed grids, where power flows are easily pre- ther anticipate new generation and build poten- dictable and significant congestion infrequent, tially unnecessary infrastructure or wait for firm zonal pricing does little to affect the efficiency of generation plans before starting the process, the resulting price signals. At the same time, larg- thereby potentially discouraging investment in er bidding zones are believed to increase liquid- new generation. Since, in the vast majority of cas- ity and competition, and zonal prices represent es, it takes much longer to plan, get approval for a more stable signal for investors and tend to be and build a high voltage transmission line than it less discriminatory for consumers. does to plan and build a wind farm or solar generating facility, transmission reinforcements are Nonetheless, many arguments used to support often significantly delayed. zonal pricing are questionable. In nodal markets, financial transmission rights can provide produc- Furthermore, in the case of wind, for example, the ers the long-term hedge required to sell their best and most efficient sites to develop projects 15 productions in forward markets to operators lo- are clustered around certain geographical areas. cated elsewhere in the system; at the same time, Therefore, unless (and until) transmission scales end-consumer tariffs can be calculated using any at the rate of the new RES generation build, the geographical granularity, i.e., averaging nodal transmission system will be increasingly con- prices, if the regulator considers this necessary strained around these areas, requiring more ac- for the sake of equity, which prevents exposing curate network representation. In this context, a some end users to higher local prices. Probably shift towards a more detailed spatial resolution the most relevant obstacles to the implementa- in the wholesale market, and therefore the use of tion of nodal pricing arise from institutional and nodal prices, may represent a useful solution. governance issues. In the United States, market 2.2.3. Bidding formats: fitting new integration is challenging because of the required necessities and resources co-ordination among different institutions and harmonisation among diverse regulations. The challenge is even greater when market integra- A large deployment of VRE requires more tion must take place across different countries, as complex bidding formats to guarantee an in the European Union (see Box 2.5) efficient economic dispatch. Bids may need to include an explicit representation of Next, we briefly discuss how, in most real-world technical constraints, as done, for instance, cases, a high penetration of variable generation in the US ISO market model. resources may aggravate the potential inefficiency of zonal pricing. As illustrated in this chapter, the variability and partial unpredictability of the increasing amounts of RES that are being deployed in power systems increases complexity and uncertainty in the short-term scheduling process.

15. Relevant examples of onshore wind include Texas, where most of the wind mills are located in the west; Germany in the north and Romania in the southeastern region of the country, close to the Black Sea. Obviously the problem of geographical location is even greater in the case of offshore wind.

52 WHOLESALE MARKET DESIGN

Box 2.5 Market integration in the European Union

Market integration is a key driver to promote overall effi- To promote efficient cross-border trades, seven European ciency and foster competition in the EU electricity market. Power Exchanges developed an initiative (price coupling of regions, or PCR) to develop a single price coupling The EU electricity market consists of a number of inter- solution to be used to calculate electricity prices across connected markets. A deep coordination among them is Europe and efficiently allocate cross-border capacity on needed to maximize the efficiency obtainable through a day-ahead basis. cross border trade between market areas, given the available interconnection capacity. It must be noted that market integration is an issue not just with regard to energy markets, but also with refer- Figure 2.12 shows electricity flows between EU markets ence to the actions TSOs have to take to maintain the in the Central Western European (CWE) region, driven by system secure. the corresponding cross-border trades, in 2016.

Figure 2.12 Annual total of physical cross-border flows in the CWE region and at the German borders in 2016 (in TWh)

NO 4.2

SE 1.1 0.3

DKW DKE 0.8 0.9 1.7

3.6 1.8

9.1 GB 1.2 0.1 PL 8.7 NL 0.3 7.7 14.8 DE 1.1 1.8 BE 4.4 6.4 CZ 11.2 5.1 7.6 FR 3.9 2 10.4 15.7 AT

1.6 CH 1.1

Source: Adapted from TenneT, 2017

Figure 2.13 shows the increase in inter TSO imbalance consisting of 11 TSOs, the largest imbalance netting co- netting – which in turn corresponds to a reduction of the operation in Continental Europe. On 2 February 2016, the balancing actions required to maintain the systems secu- French TSO RTE joined the IGCC. As shown in Figure 2.13, rity – made possible by a deeper inter TSO cooperation. inclusion of this large Load Frequency Control Block has To be more precise, figure 2.13 shows the results of the resulted in a significant increase of the netted volumes. International Grid Control Cooperation (IGCC) currently

53 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

2 Figure 2.13 Monthly volumes of netted imbalances for all IGCC members for the imbalance netting process 2012 2013 2014 2015 2016 500

400

300 GWh 200

100

0 July July July July July May May May May May June June June June June April April April April April March March March March March August August August August August January January January January January October October October October October February February February February February December December December December December November November November November November September September September September September

AT BE CH CZ DE DK FR NL

Source: Adapted from TenneT, 2017

Short-term markets are nothing but tools that aim we find two major approaches to the bidding for- to help both market agents and system operators mat design: 1) the more complex ISO bidding for- approach this objective in the most efficient way mat (used in the United States) (see Box 2.6), and by optimising economic risk management for the 2) the simpler PX format (used in the European former and achieving short-term reliability and Union). security levels, a responsibility of the latter. ISO bidding format and increasing penetration In this context, the format of the bids that mar- levels of RES ket agents can use influences the ultimate mar- In the US, in principle, increasing the penetration ket outcomes. The alternatives and the flexibility of RES need not significantly impact the com- made available to market agents when building plexity of the format applied to existing resourc- their bids is central to allowing them to reflect es since the model is already based on complex their true costs and the physical constraints of and detailed (multipart) offers. That being said, their generation units on the energy market. Bid- enhanced modelling software, made possible by ding formats differ from one market to another, developments in computing technology, has al- and they can range from a simple price-quantity lowed a progressive increase in modelling detail hourly bid (indeed, 24 price-quantity hourly bids) (O’Neill et al., 2011). to a complex declaration of technical and economic constraints. Where the model has to be further refined is in integrating new resources, particularly those While the optimal design of bidding formats has with characteristics that differ from the status been an open question since the creation of elec- quo. This calls for a larger number of bidding tricity markets, it has recently been revealed as a formats tailored to the needs of these new re- critical subject that needs to be revisited. Again, sources. For example, in the relevant case of de-

54 WHOLESALE MARKET DESIGN

Box 2.6 Bidding formats in the US ISO market: multipart offers

Independent system operators (ISOs) require generators ling a conventional generator, with the aim of defining the to submit multipart offers aimed at representing the de- economic dispatch in the most efficient way. tailed operational (and opportunity) costs and also the Other bidding formats have also been implemented for technical constraints of their generating units. Table 2.2 different types of resources, such as multistage resources lists several typical offer components used while model- (combined cycles), intermittent generation, pumped-storage hydropower units and other storage units.

Table 2.2 Typical multipart offer structure in ISO markets

Operating costs Technical constraints

Energy offer curve MW, $/MW Economic min MW

Piecewise linear or stepwise linear function Economic max MW with multiple MW/Price pairs Ramp rate MW/hour

No-load offer $/hour Min/max run time hours, min

Min downtime hours, min

Start-up offer $ Notification time hours, min

Available for different types of start-ups Cooling time hours, min (hot/intermediate/cold) Start-up time hours, min

Note: Small storage, typically electrochemical storage, is in many cases only allowed to provide secondary frequency control reserves (regulation).

mand response, as pointed out, for example, by lar demand. This was simpler to predict in a pow- Liu et al. (2015) “the bidding system does not al- er sector with stable and predictable dispatches, ways provide a mechanism as an alternative to but it becomes quite risky when the high pene- the price-quantity bid format for consumers to tration of variable technologies can unpredict- express their willingness to adjust consumption, ably alter the price. particularly in response to price signals”. For in- The main obstacle to creating more complex bid- stance, an industrial consumer may need a cer- ding formats is the computational complexity of tain number of consecutive hours of supply, but the associated optimisation model. This limiting is willing to shift this period according to the mar- factor is not as relevant to the EU model. ket price. This willingness cannot be expressed as a simple quantity-price bid during each hour; it The EU power exchange approach: the complex- needs a complex bid format. ity of prioritising simplicity A similar difficulty is related to the consideration As pointed out in the beginning of this chapter, of storage technologies, as uneconomical sched- most European markets were designed with the ules can often occur using current bidding for- explicit aim of reducing – as much as possible mats. For example, they often have to define ex – the intervention of the system operator. Con- ante (before knowing the market prices) both the current aims include: releasing market agents periods when they want to act as producers and from the need to buy and sell through a com- those when they want to procure energy as regu- pulsory pool (a market model that centralises all

55 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

trades and covers its costs through fees to mar- The main goal of the new bidding formats is to ket agents), decoupling the format of bids from allow PX participants to hedge their risk against a 2 plants’ technical characteristics (costs and con- wrong estimation of market conditions (and thus straints), and postponing as much as possible the minimise the risk of being scheduled in ways that market gate closure.16 imply additional costs), while at the same time ensuring that auctions are transparent and com- Day-ahead markets in European power exchanges petitive. This allows market agents to build strat- were originally, and at least in theory, envisioned egies that somehow internalise the potential im- as simple electricity auctions, where most agents pact of the physical and economic constraints of would submit price-quantity offers and bids. their portfolios in a more effective way than just The reality is not as straightforward. For simple through simple price-quantity bids. bids to be utilised in a context where the laws of Early on, the use and impact of these complex physics play such a strong role, market agents and block bids was rather irrelevant. A simpler must anticipate the resulting dispatch and inter- approach was thought to ensure greater trans- nalise all operational costs and constraints, a task parency of clearing results. Today, the uncertain- that, to be properly fulfilled, requires that market ty that market agents face when building their conditions be possible to predict (the lower the bids has significantly increased, mainly due the size and diversity of the portfolio of plants each higher shares of variable renewable energy, and market agent manages, the harder this task). therefore the use of complex and block orders is The increasing penetration of RES has increased on the rise (see Box 2.8). uncertainty in the dispatch, making accurate It is often argued that an increase in uncertainty predications more difficult. This calls for bidding and a lack of more accurate bidding formats are formats that allow for greater complexity and re- not necessarily problems if sufficiently liquid and duce the burden for market agents. In response, thus efficient intraday market sessions follow the as seen in the Box 2.7, each PX has progressively day-ahead market auction. Certainly, as we have incorporated so-called block orders and semi- seen, the role of intraday sessions is relevant to complex orders, (rather than the multipart for- allowing market agents additional resources to 17 mats used in the US ISOs). readapt the schedules resulting from the day- ahead market clearing, but this argument is not Box 2.7 Bidding formats in European without challenges. power exchanges On one hand, the algorithmic complexity of the European power exchanges use the EUPHEMIA mar- EU clearing mechanism does not handle large ket-clearing algorithm in their day-ahead market to amounts of block bids well, the reasons for which support the different bidding formats used by each will be reviewed in the next section. On the oth- market operator. Simple bids can be used in all bid- er hand, market agents may in the future need ding areas, but depending on the market operator, to combine increasing amounts of block orders agents can represent some operational constraints in order to reliably represent their characteris- and economic conditions using block orders or tics and hedge against different potential sce- complex conditions (see Appendix: EUPHEMIA, the narios (see Appendix). This has led the current pan-European market clearing algorithm). approach to be questioned since the number Block orders basically entail an all-or-nothing con- of block orders can become impractical. In oth- straint, representing the willingness of an agent to er words, properly hedging against all potential sell/buy some amount of electricity at a given price, market outcomes may require bidding on an ex- but only if this amount can be sold/bought fully. Ad- tremely large number of block orders. For this ditionally, several block orders can be combined to reason, one of the root principles of EU PX is be- create more complex products. ing reconsidered: the key idea is whether the algorithm performance may be improved by us-

16. These and related aims are particularly apparent since the implementation of the New Electricity Trading Agreements. 17. In some cases, the block and complex conditions were already implemented in the initial market design.

56 WHOLESALE MARKET DESIGN

Box 2.8 Empirical evidence of the growing role of complex conditions and block bids in EU power exchanges: the case of Spain

Figure 2.14 illustrates how, even in a power market with a It can be argued that when the amount of offers reject- relatively large amount of hydropower reservoir resourc- ed by the algorithm becomes large, the efficiency of es, the impact of complex conditions (in this case the the market results can be put into question. A market minimum income condition, or MICa) increased alongside clearing based on multipart offers (as in the US ISOs) wind penetration (Vázquez et al., 2014). The graph shows probably would not reject (all of) those bids and would how, until 2005, the Spanish market cleared and behaved minimise the overall cost of supply. Indeed, multipart as a simple auction; the number of offers rejected during offers allow producers to properly represent the tech- the clearing process because of the MIC condition was nical constraints and the cost structure of the different types of plants. Conversely, block bids seek to allow close to negligible. Starting from 2006, however, due to for only an extremely simplified representation of the increased penetration of wind resources and the con- those constrains, precisely with the risk of completely sequent increased variability in the daily dispatch (more rejecting the offers of some plants that, if they could cycles and ramps for thermal units), the MIC conditions offer multipart offers, would have been cleared for an started being activated much more frequently. Many bids efficient subset of hours. In recent years, the number were rejected since the total income did not reach the of complex conditions used by agents has remained specified minimum. relatively constant (PCR-ESC, 2015).

Figure 2.14 Evolution of the activation of the minimum income condition and wind production

Rejected energy through MIC activation (peak hour)

30 20

GWh 10 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Daily Wind Production

300 200

GWh 100 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Source: Adapted from Vázquez et al., 2014 a. As defined by the Spanish market operator, OMIE “sellers may include, as a condition governing the electricity sale bids they submit for each production unit, that the bid in question is only to be considered submitted for matching purposes if the seller obtains a minimum income. The minimum income required shall be expressed as a fixed amount in €, and as a variable amount expressed in €/MWh” (Omel, 2001).

57 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

ing resource-specific bidding formats that would These characteristics hold only under certain as- only require one (multipart) offer per resource, in sumptions, including the absence of economies 2 line with the ISO approach already reviewed. of scale in generation and the absence of “lumpy decisions” (Hogan and Ring, 2003), stemming, As pointed out at the beginning of this section, for instance, from “lumpy costs” (e.g., start-up one alternative – recently supported in PCR-ESC costs) or “lumpy constraints” (e.g., all-or-nothing (2015) – is to resort to multipart offers. This ap- commitments or minimum outputs).19 In reality, proach involves greater complexity than blocks electricity markets present multiple and unavoid- or complex orders, but it is preferable if the multi- able lumpy decisions. part offer is able to replace multiple blocks. How- ever, the EUPHEMIA provider has pointed out In the US ISO model, multipart offers contain that including such a bid would involve a signifi- both lumpy costs and lumpy constraints. In the cant change in the market design and pricing and EU contexts, simple bids do not introduce “lumpy clearing rules, a design element which will be an- decisions” but block bids and other complex con- alysed in the next section. ditions do. Thus, it is mathematically impossible to find uniform prices that support the wel- 2.2.4. Pricing and clearing rules fare-maximising solution (Scarf, 1994). Practical implementation of marginal pricing More complex markets require modifications therefore includes modifications that accord to clearing and pricing models. more or less priority to the two above-mentioned Of the two main approaches to pricing rules objectives. Two real-world examples follow. and clearing – those of the US ISO (dispatch- •• Optimal-dispatch-based (marginal) prices: The based pricing plus discriminatory pricing) and volumes accepted on the market are those of the EU PX (price-based dispatch) – the best the welfare-maximising solution (optimal dis- option might be some point in between. patch). However, a discriminatory pricing rule must be applied to guarantee cost-recovery. The key economic theory at the basis of electric- Marginal prices (e.g., hourly) are calculated, but ity market designs is marginal pricing remunera- some agents may have to pay/receive an addition for generation. Under this principle, at each tional lump sum (uplift) to avoid cleared bids point in time, electricity is supposed to be valued that would be at a loss if only marginal prices at the marginal cost of producing (or not con- were considered. This approach sacrifices effi- suming) an additional unit of energy. Ideally, this cient price signals for short-term operational has two benefits: efficiency. •• Uniform-(marginal)-price-based dispatch: A •• First, the marginal price supports the welfare uniform pricing rule constraint is imposed; maximising solution.18 That is to say, accepted i.e., all transactions in a given period (e.g., an bids are sufficiently compensated, and rejected hour) are settled at the same price. The level bids are not profitable at the marginal price. of (uniform) market prices in the various hours •• Second, settling the transactions of all agents of the day must therefore be such as to include at the same price (uniform pricing) sends an ef- – not necessarily in a perfect way – the impact ficient signal for bidding true costs and for op- of lumpy constraints and costs. This price con- timal investment over the long term (the main straint requires that the market solution deviate argument in favour of the energy-only market from the most efficient (social welfare maxim- approach to ensure resource adequacy). ising) dispatch.

18. Ventosa et al. (2013) define welfare maximisation as the “overall minimisation of system costs while respecting reliability and environmental objectives and constraints.” See the same source for the economic formulation of social welfare in the power sector context. 19. More exactly, the problem is the presence of non-convexities in the optimisation problem (the maximisation of the social welfare). These non-convexities arise, for instance, from the discrete “jumps” in the cost function of thermal plants, which have a constant variable cost, but only if a fixed start-up cost is incurred at the beginning of the commitment period.

58 WHOLESALE MARKET DESIGN

Uplifts are unavoidable elements of the dis- Box 2.9 Illustrating the need for uplifts patch-based pricing system required to support A plant submits an offer to the independent system the welfare-maximising dispatch. The underlying operator (ISO), including its variable cost per mega- problem is the fact that not all costs incurred are watt hour (MWh), along with other costs such as embedded in uniform market prices, which alters start-up and no-load costs. Once the plant has been the correct signals.20 committed, its variable costs and the variable costs of other committed resources are the basic drivers of The problem with energy uplifts the marginal price and the dispatch level of the unit. Resorting to uplifts not only prevents prices from

Absent other constraints (such as ramps), when the reflecting the full cost of the serving load, but its price is greater than the variable cost, the resource associated cost has to be somewhat arbitrarily al- is dispatched at the maximum output, and when the located to energy consumers. Currently, the up- price is less than the variable cost offer, the plant is lift calculation involves only the generation side either shut down or operated at its economic min- of the market, taking advantage of the fact that, imum (depending on whether it makes sense eco- because of demand-side inelasticity, it can be de- nomically to de-commit the resource). termined ex post and then socialised. In a context of active demand participation, the uplift alloca- This setting illustrates two clear situations where the tion should be incorporated in the market-clear- market price may not be sufficient for the unit to re- ing process to be consistent with the demand cover its short-term operation cost: side as well. −− If the unit is marginal, then the price is equal to its variable cost, and the unit will not recover its start- As already mentioned in the previous section, up and no-load costs. one of the effects of RES penetration in electricity markets is the increasing need for complex If the unit is committed but dispatched at the eco- −− bidding formats, which aggravate the pricing nomic minimum, then the price is lower than its vari- problem. In the ISO context, this problem has re- able cost (since the unit is not marginal), and the cently attracted a lot of attention (see Box 2.10), variable cost is not fully recovered through the mar- and ISOs are undergoing design improvements ket price. aimed at reducing uplift. Proposed solutions at- In these cases, and others, the resource needs to re- tempt to internalise all costs in market prices as ceive an uplift payment in order to be dispatched at much as possible – in other words, to approach the optimum level, as determined by the ISO, with- uniform pricing. This represents the increasing out losing money. importance of short-term market signals needed for efficient generation investment and the development of demand-side resources (Hogan, Next, we analyse the challenges involved in each 2014). particular approach and possible solutions being explored today. Pricing and clearing in the European Union: Uniform-price-based pricing approach Pricing and clearing in the United States: The optimal-dispatch-based pricing approach Although one of the claimed objectives of the EU- PHEMIA clearing algorithm is the maximisation of In the United States, the ISOs first calculate the social welfare, the algorithm considers two ma- optimal dispatch and then compute prices based jor rules to be fulfilled that condition the clearing on the marginal cost of the system. On the basis process and thus the corresponding associated of these prices, uplifts are calculated to compen- social welfare: sate generators incurring costs above the revenue earned through market prices. See Box 2.9 for an •• The single marginal pricing rule entails that uni- illustrative example based on Bresler (2014). form market prices (without uplifts) must suf- fice to compensate all accepted bids.

20. This is instrumental, in theory, to leading long-term market decisions towards the optimal capacity expansion (see Herrero et al., 2015).

59 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 2.10 Uplifts: an increasingly relevant problem in some US systems

The Federal Energy Regulatory Commission (FERC) initi- system constraints make necessary subsequent redis- 2 a ated a docket to improve price formation among US inde- patches commanded by the ISO (FERC, 2014b). pendent system operators (ISOs). This docket identifies While uplift cannot be completely eliminated, it can be uplift as a major problem (FERC, 2014a): largely reduced with appropriate market design, as stat- “Use of uplift payments can undermine the market’s ed by Pope (2014): ability to send actionable price signals. Sustained “Uplift is a symptom rather than a cause of price patterns of specific resources receiving a large formation problems, though, and efforts to improve proportion of uplift payments over long periods of pricing should focus on correcting the causes”. time raise additional concerns that those resources are providing a service that should be priced in the Such causes are varied, and the solutions will affect very market or opened to competition”. diverse parts of the market, but, undoubtedly, a key element is the method used to compute prices. A widely Uplifts can occur for many reasons; the three primary debated solution is the so-called extended location- ones in ISO markets are: 1) the operating costs of some al marginal pricing, which in the literature is known as resources are not reflected in prices (the case reviewed Convex-Hull pricing (Gribik et al., 2007). Essentially, this in Box 2.9); 2) inflexible resources, such as the so-called method would produce hourly prices that better reflect block-loaded units, are committed and 3) unmodelled the full cost of producing electricity (e.g., internalising the start-up and no-load costs).

a. Inefficient price distortions become particularly evident for the case of block-loaded (or fixed-block) units, which are those that only operate economically at full load. Fast-start gas turbines generally lie in this category. When these units are committed, other units have to be dispatched down to accommodate the full output of block-loaded units. If block-loaded units are treated as non-dispatchable by enforcing a minimum economic output constraint, they cannot set marginal prices and, therefore, require uplift payments to recover their operational costs.

•• Simple bids are given preferential treatment offs, and in the European context, uniform pricing (over complex and block bids): if the market is considered an objective worth the loss in short- price is above a simple bid price, the bid always term cost efficiency (or, more generally, welfare has to be fully accepted (in the terminology maximizing). One of the advantages of uniform used by EU PXs, “simple bids cannot be para- pricing is that demand and generation interact on doxically rejected”). the market in equal terms, and it is not necessary to define rules to allocate uplift that would inevi- In the presence of “lumpy decisions”, these two tably send inefficient signals. rules make the market-clearing results deviate from those obtained in the US ISO model. Box 2.11 The European approach does, however, present (based on Olmos et al., 2015) offers an example some problems that compromise the sustainabil- of how these two rules may affect market results ity of this model, namely: 1) the complexity of and divert them from those obtained in the US the algorithm limits the amount of complex and ISO model. The example is based on a block-load- block orders that can be handled in practice; 2) a ed plant. As was noted in Box 2.10, inflexible re- certain lack of transparency in the algorithm be- sources are among the major reasons why uplift ing used; and 3) the inability of some units to set is needed in the United States. the market price. The European market-clearing mechanism makes Computational complexity the short-term dispatch of generation units devi- One of the major practical downsides of the EU ate from the welfare-maximising dispatch. As ar- approach is that integrating clearing and pricing gued before, however, this is a matter of trade- in one step is inevitably a problem harder to solve

60 WHOLESALE MARKET DESIGN

Box 2.11 Illustrating the difference between the US and EU pricing and clearing

Consider a one-hour market setting and an inflexible block type of plant. To illustrate how lumpy decisions are dealt Figure 2.16 Result with the ISO model approach with in the EU and US approaches, it is useful to consider the market setting represented in Figure 2.15.

Within the generation bidding curve (in blue), we introduce a block order (dotted line). This is an all-or-nothing type of

bid that is either completely matched or completely reject- $/MWh ed. The marginal price is the cost of supplying a marginal Welfare increment in demand, and it corresponds to the variable cost of the margin unit, i.e., the unit in the position of sup- Indivisible plying such marginal increment. Because of the block-order Pa condition, the bid represented with a dotted line will never be marginal, for it cannot “marginally” supply an additional MW MWh of demand due to its inflexible nature.

In the EU Price Coupling of Regions (PCR), EUPHEMIA Figure 2.15 Pricing and clearing in the presence also seeks to maximise the welfare, but at the same time of lumpy bids it has to comply with the two rules previously outlined: the uniform market price (no uplifts allowed) and the full acceptance of simple bids if the bid is below the market price. Since the indivisible block needs a price equal to or above its offer to be accepted, it is not possible to par-

$/MWh tially accept any simple bid below that price. As a conse- Indivisible quence, the algorithm rejects the block order. This leads All-or-nothing block to the dispatch depicted in Figure 2.17: the most expensive bid cannot be accepted because it exceeds the price of the demand bid. Therefore, only part of the demand will be supplied, and the price is set by the partially accepted demand bid. As can be observed by comparing MW Figures 2.16 and 2.17, welfare is much lower with the second clearing approach. In the United States, the welfare-maximising dispatch requires reducing the production of a cheaper flexible plant to make room for the indivisible bid. Since the indivisible Figure 2.17 Result with the EU PX model approach bid cannot be marginal, the price in the market is set by a lower-price bid (Pa in Figure 2.16). With such a price, the block-loaded plant would not recover its short-term productions costs. Thus, the need for an uplift. In the figure, the orange horizontal line represents the price of the de-

mand bid (a single inflexible demand bid is considered). As $/MWh explained in Ventosa et al. (2013), the social welfare can be Welfare represented as the area contained between the cleared demand curve and the cleared offer curve (green area in the Indivisible figure 2.16).

Source: Olmos et al., 2015

61 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

than the economic dispatch problem followed by “The effect of defining an order as a block is an ex post price (van Vyve, 2011). that the order cannot then be a full price mak- er. Rather, block orders may impose a bound on 2 This led some PXs in the precoupling era to lim- the range of prices possible while the price be- it the amount of block orders that could be used. ing set would still need to come from the simple Further developments in computing and solving order or complex order curves. This is because techniques allowed for an increase in the num- the decision to execute the order is an integer ber of block bids that could be handled, as well decision (i.e., the order is executed or not ex- as the introduction of new block bid formats. The ecuted) and the decision on whether to accept increasing use of these formats is, however, threat- a block occurs before the price determination ening the sustainability of this approach. sub-problem. The bound created by the last ac- Transparency cepted block order would function to affect the The EUPHEMIA algorithm has to deal with a diffi- price (by limiting possible values) but could not cult combinatorial problem: deciding which com- directly set this price”. plex and block orders are accepted and which are “This was discussed with the PCR ALWG rep- rejected. How EUPHEMIA accepts and rejects the resentative, APX, who confirmed that without orders is reason for debate. It is difficult to justify the blocks setting the price, the price could only why some orders are rejected by the EUPHEMIA be set by other price makers, i.e., simple orders algorithm when the prices would not allow them or complex orders, or the price indeterminacy to be accepted at a loss (these are the paradox- rules of EUPHEMIA”. ically rejected blocks, or PRBs; for example, the indivisible block in Figure 2.17 is rejected even if The European approach requires simple enough the clearing price is higher than its offer). In PCR- bids to ensure the price is representative of sys- ESC (2015), the Market Parties Platform (MPP) tem costs. In a scenario where most bids are pointed out that: “There may exist false PRBs: re- block orders, the market price will not accurately jected in-the-money blocks that could have been indicate the marginal cost of meeting the load. accepted and result in a better (higher welfare) Trends in US and EU pricing rules: solution. MPP asks for more transparency on op- towards a hybrid solution? timality, to prove the absence of false PRBs”. The problems highlighted in US and EU pricing This is linked to the computational complexity rules share the same primary causes and could problem, which has led to a complicated clear- therefore benefit from similar solutions. To some ing algorithm. The public documentation of the extent, the novel pricing methods discussed market-coupling algorithm (PCR PXs, 2016) is in the US ISO context can be seen as a middle not completely detailed. The joint response of ground between the reviewed US and European ACER and the Council of European Energy Regu- approaches. lators (ACER-CEER, 2015) to the European Com- While the trend in the United States is to advance mission’s Consultation on a new Energy Market towards a hybrid solution, this possibility is gen- Design stated: “We would particularly like to see erally ignored in Europe, where current efforts clearer rules and greater transparency around concentrate on improving the computational the market coupling algorithm (EUPHEMIA)”. performance of EUPHEMIA in order to cope with Not all units can set the market price the increasing number and complexity of the bids According to the authors’ best interpretation of that are used. However, it seems that more rad- the public description (PCR PXs, 2016), the Euro- ical solutions may be needed to decidedly solve pean pricing rule also suffers one of the pricing the complexity issue. In this respect, three long- problems we have seen in the US model: inflex- term solutions are pointed out in PCR-ESC (2015): ible block bids cannot set the market price. This is confirmed by the thorough analysis carried out by Eirgrid et al. (2015), who stated:

62 WHOLESALE MARKET DESIGN

•• Reduce the amount of block types and other Although operating reserve requirements are a complex products allowed per participant and regular requisite in electricity systems, defining market (bidding zones). these requirements for security reasons is a clear •• Reduce the range of products treated in EU- but well-accepted intervention of the central PHEMIA. planner (the system operator in this case). The •• Relax the uniform price requirement (accept system operator (See Box 2.12) needs to ensure that the result has more than one price per bid- the availability of a certain level of operating re- ding zone and time period). serves to tackle unpredictable short-term events and avoid forced curtailment or cascading fail- Among these, only the last one really tackles the ures in the system. root of the problem. In this regard, one of the alternatives being discussed is proposed in van Two interdependent products Vyve (2011), whose model resembles the ISO The values (prices) of reserves and energy are approach in that it uses the welfare-maximising mutually dependent. The fact that generation solution and compensates committed units at a plants can offer one or the other product links loss through uplifts. the value of reserves with the opportunity cost of providing energy, and the other way around 2.2.5. Rethinking reserve requirements (Stoft, 2003). and procurement Because of this interdependence, if the objective is to provide accurate short-term price signals, it System operators need to implement new must be taken into consideration that both the solutions to improve the reserves supply requirements and the procurement mechanism function in such a way that they are priced (co-optimised or not) for operating reserves mat- according to their real value. ter. At the same time, energy and reserve markets need to be properly connected in order to As discussed by Hogan (2013), in the face of effi- allow the former to reflect the actual value of ciency, this calls for two relevant conditions in the the latter. market design: •• A proper definition of the operating reserve de- Short-term energy markets produce dispatch in- mand curve (ORDC) is needed. Roughly speak- structions that are very close to the actual deliv- ing, the ORDC needs to account for the real ery of energy. However, the final power plants’ value that any amount of reserves has for the output or the demand consumption can deviate system. from these commitments. This calls for last ad- •• Energy markets have to reflect the opportunity justments during real-time operation, which can cost of providing reserves and potential scar- either be made automatically or by command of cities in the reserve product in such a way that the system operator using different types of op- the prices of the two products maintain a stable erating reserves (also referred to simply as re- and transparent linkage. serves). We next analyse these two desirable require- An operating reserve is defined by NERC (2015) ments. as “that capability above firm system demand re- The operating reserves demand curve (ORDC) quired to provide for regulation, load forecasting System operators usually define a minimum con- error, equipment forced and scheduled outages tingency requirement for reserves. Below this and local area protection” and is typically divided minimum level of reserves, the load would be under primary, secondary and tertiary frequency curtailed to ensure that the reserve target is met. control. Usually, these quantity requirements are procured through market mechanisms. The demand

63 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 2.12 TSO’s actions to ensure the availability of reserve resources 2 One of the actions undertaken by the system operator in Similar functions in US systems are carried out by the ISO the ancillary services markets is to buy (sell) power from through the reliability unit commitment, where the ISO (to) specific power plants to ensure the availability in the adjusts the day-ahead market’s schedules to take into ac- system of sufficient spare resources capable of rapidly re- count its load forecast and all security constraints, includ- sponding to dispatch orders in real time, so as to maintain ing the availability of adequate backup resources. the system balanced (see Figure 2.18 for sample).

Figure 2.18 Sample of TSO’s actions to ensure the availability of reserve resources

Day-ahead (+intraday) results

PMAX PMAX

PMIN PMIN PU1 PU1 PMAX TSO ancillary PMAX Sold by TSO services market PMIN PMIN PU2 PU2

PMAX PMAX

PMIN PMIN PU3 PU3 Purchased by TSO UP3 not running at all Enough fast Post day-ahead Post TSO’s ancillary No reserve response energy market service’s market reserve

PU = Production Unit

curve in such a market is completely inelastic,21 This minimum requirement and its value has been i.e., a demand bid is put for the entire quantity long debated in the literature. Stoft (2003), for (equal to the contingency requirement) at a price instance, says in this respect: “Operation reserves equal to the value of lost load (VOLL). This means requirements, though based on sound princi- that the willingness to pay for each MW of reserve ples, are rules of thumb that vary from one con- up to the minimum contingency level is set to the trol area to another. Is it possible that each mega- VOLL, while above this level, the marginal value watt of operational reserve is worth U.S.$10 000 of any additional MW is typically set to zero. The up to the requirement, but the next megawatt is price for the operating reserve is then set accord- worth nothing? Is it possible that there is no limit ing to supply bids from market agents. The in- on what the system operator should pay to meet elastic demand bid on the reserve market reflects the last megawatt of the requirement?” the priority that system operators assign to secu- Resorting to first principles, probably what is rity over economic efficiency, potentially leading more questionable is the zero value above the to over-procurement of operating reserves. minimum contingency level. As described by Ho-

21. An inelastic demand does not change its consumption behaviour depending on the price. In the reserve market, the price almost never reaches the VOLL level. Thus, in normal operations, reserves are procured up to the contingency level, no matter if their price is 10 or 1 000 USD/MWh.

64 WHOLESALE MARKET DESIGN

gan (2014), whenever there is in real time a forced in most US ISOs’ day-ahead markets. However, load curtailment, the value of having available ad- the co-optimisation is not always carried out in ditional amount of reserves (above the minimum the real-time market segment, where it is proba- contingency requirement) would correspond to bly more important. the VOLL during that period. This is the case in ERCOT, where the market in- At any previous time frame, such value would cludes co-optimisation of energy and reserves to be probabilistic (based on the expectation of re- clear the day-ahead market, while the real-time ducing the potential curtailment). In other words, market does not. This lack of co-optimisation has the value of an increment of operating reserves led to the connection between the ORDC and the would be the VOLL, but now multiplied by the energy market being implemented through the probability that these additional reserves help re- second-best approach – that of using a price ad- duce the curtailment. der to the energy price. This price adder aims at replicating, approximately, the outcome of co-op- This leads to an ORDC that has the shape of a timisation (Hogan and ERCOT, 2013). loss of load probability (LOLP) function, at least above the minimum contingency level, as shown Figure 2.20, from ERCOT (2014), illustrates how in Figure 2.19. the adder works and how it depends on the reserves available and the energy market price. The effect of using such a curve would increase On the one hand, the price of reserves depends the price of reserves, and since the energy and on the quantity of reserves procured: the higher reserve prices should be “connected” in properly the price, the lower the quantity, with the price functioning markets, the energy price would also being USD 9,000/MWh for reserve levels below increase. This connection is further explored next. 2,000 MW. On the other hand, the adder that Designing the market to allow energy prices “connects” the price of energy with the price of to reflect the opportunity cost reserves will depend on the gap between the pric- of providing reserves es of both products, and will therefore depend The co-optimisation of energy and reserves is the on the level of reserves and the price of energy. textbook solution that results in the appropriate This way, for a reserve level below 2,000 MW and valuation of energy accounting for the oppor- an energy price of USD 9,000/MWh, the adder tunity cost of providing reserves (and the other would be close to zero (for the energy price is way around). This is the approach implemented already reflecting the non-served energy price).

Figure 2.19 Operating reserves demand curve

12,000

VOLL 9,000

8,000

6,000

4,000

2,000 Price of Reserves (USD/MWh) of Reserves Price 0 0 2,000 4,000 6,000 8,000 Available reserves (MWs)

Source: Adapted from ERCOT, 2014

65 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 2.20 Price adder as a function of the energy However, for a reserve level below 2 000 MW and price for different reserve levels an energy price of USD 2 000/MWh, the adder 2 will be USD 7 000/MWh. $9,000 4,000 It is important to note that the value of the adder $8,000 3,500 depends on the level of reserve available. Indeed, $7,000 3,000 the adder’s goal is to properly price the energy $6,000 2,500 sold in the day-ahead market. It requires taking $5,000 2,000 into account the system’s scarcity, which in turn $4,000 1,500 depends on the available reserve margin once ex- $3,000 pected demand is satisfied. Therefore, the value 1,000 $2,000 of the adder grows at the shrinking of the avail- 500 $1,000 able reserve to reflect the higher risk of having to $- -MW unintentionally disconnect the load. $ Energy $ Adder Reserves MW Box 2.13 summarises the results of the backtest-

Source: Adapted from ERCOT, 2014 ing (involving years 2011 and 2012) carried out in Texas to assess how energy prices would have changed had this operating reserve demand

Box 2.13 Techno-economic analysis of the impact of implementing the ORDC in Texas (Backtesting for years 2011 and 2020)

“The Public Utility Commission of Texas (PUCT) request- “The back cast analysis shows that the energy-weighted ed that ERCOT perform a back cast of an interim pro- average energy price increases over a range of $7/MWh posal that intended to be a more appropriate method of to $26.08/MWh in 2011 and $1.08/MWh to $4.5/MWh in pricing reserves. The back cast approximated the pricing 2012. This range results from different parameter settings outcomes and estimated what the market impacts may that were used in the back cast. The back cast results for have been if the solution devised had been in place for the the average energy price increase with minimum contin- years 2011 and 2012”. gency levels (X) of 1375 MW and 1750 MW are presented in Table 2.3. At the minimum contingency level, scarcity prices achieve the maximum allowed value”.

Table 2.3 Energy-weighted average energy price adder (and online reserve price)

Energy-weighted average price Energy-weighted average price increase with X at 1375 MW (S/MWh) increase with X at 1750 MW (S/MWh) VOLL 2011 & 2012 2011 & 2012 2011 2012 2011 2012 combined combined

$5000/MWh 7.00 1.08 4.08 12.03 2.40 7.28

$7000/MWh 11.27 1.56 6.48 19.06 3.45 11.35

$9000/MWh 15.54 2.05 8.87 26.08 4.50 15.42

Source: Hogan and ERCOT (2013).

66 WHOLESALE MARKET DESIGN

curve been implemented with a co-optimisation 2.3. BALANCING MARKETS of energy and reserves (Hogan and ERCOT, 2013).

In the European Union, as introduced above, Properly designing balancing markets there is a separation between energy markets is essential to ensure that: and reserves markets. As briefly illustrated in the 1) Accurate incentives for flexibility are next section, they are sequential markets where offered agents can (and indeed do) withhold capacity from the energy market (though including in 2) All resources can effectively participate their bids the opportunity costs of providing rein offering their flexibility potential to serves in the ancillary service market) to provide system operators. reserves.22 This way, one market ideally reflects the opportunity cost of providing the other ser- In the previous section, we discussed a multitude vice. Unfortunately, in real life, the information is of issues related to day-ahead and intraday mar- not perfect, and undesired deviations from an ef- kets and introduced the concept of gate closure. ficient outcome are likely. After gate closure, the system operator must en- Co-optimisation, although representing the first- sure overall system security and stability in real based approach, would embody a major institu- time, which calls for an instantaneous (in the or- tional challenge in the European Union today and der of seconds) matching of electricity supply and so is not a realistic alternative. demand. Although RES obviously add complexity to this challenge, practical experiences have The main problem with the EU approach is relat- shown how power systems can be safely oper- ed to the timeline of the procurement of reserves. ated in real time with large levels of renewable While there are frequent markets to procure and production, provided that the system is prepared sell energy, reserve markets are not so frequent. for this. The question today, however, is how to Typically, reserves are procured in a single ses- ensure this real-time security in the most efficient sion, whose lead time varies from country to manner to better integrate large volumes of RES. country (month ahead, week ahead, day ahead, In this respect, there is a total consensus on the etc.). This lack of frequent markets for reserves is key role of properly designing balancing markets a major drawback, preventing the effective con- and mechanisms. tribution and participation of many RES and distributed energy resources (DER). Furthermore, Technological innovation allows variable renew- procuring reserves exclusively in the long term able energy technologies to provide ancillary ser- may be inefficient because it does not allow for vices, ultimately contributing to system flexibili- the complementarity between reserves and en- ty and reliability (See Box 2.14). A well-designed ergy to be properly considered. balancing market should give incentives to all types of resources, including conventional generation, demand and RES, to offer their flexibility potential to system operators. Nonetheless, the regulation governing the provision of these services, hampers RES generators and DER from delivering them. This section characterises the essential elements of balancing services and markets with a direct impact on the adequate integration of RES and DER resources, also highlighting best practices in their design. The section is divided into three parts: 1) the imbalance responsibility and settle-

22. In some EU markets the sequence is the other way around – reserves are procured before the energy.

67 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

the EU context, the unit for settling the total net Box 2.14 Utility solar PV plant’s ability to provide imbalances is known as the balancing responsi- ancillary services to the grid 2 ble party (BRP). The imbalances from the pro- The technical capabilities of utility-scale PV plants gramme declared at gate closure are measured to provide ancillary services have been known for and charged to these BRPs. a number of years. Such features have also been recently proven in commercial settings. A recent In the words of ENTSO-E (2014), “in a liberalised study on a 300 MW PV plant carried out by CAISO, market, the market players have an implicit re- NREL, and First Solar shows that the solar unit per- sponsibility to balance the system through the formance was able to meet the frequency regula- balance responsibility of market participants, the tion response usually provided by natural-gas-fired so called Balance Responsible Parties or BRPs. In peaker plants. This is another demonstration of the this respect, the BRPs are financially responsible contribution that utility-scale solar farms, equipped for keeping their own position (sum of their in- with advanced inverters and software controls, can jections, withdrawals and trades) balanced or to offer to smooth out grid fluctuations, enhance sys- help restore system imbalance over a given time- tem flexibility and reliability, and reduce needs in frame”. spinning reserves. The allowed aggregation level for measuring im- Source: CAISO, 2017 balances conditions the quantities that are subject to imbalance charges (since imbalances within the BRP in opposite directions compensate ment, 2) the definition of the balancing products each other). The criticality of the aggregation lev- and 3) the design of markets for procuring bal- el will be explained later, as it is very much related ancing reserves and their pricing and remunera- to the adopted pricing scheme. tion schemes. Imbalance settlement 2.3.1. Balancing responsibility The other element that completes the imbal- and imbalance settlement ance charges is the price to be applied. There are two major price schemes that can be applied to Dual imbalance pricing does not reflect BRPs for their net imbalances: the single- and du- imbalance costs and therefore distorts the al-pricing schemes. real-time price signal. In the single imbalance pricing scheme, imbal- How to define balancing responsible parties ances are always settled at prices representing (BRPs) is a contentious issue when coupled the procurement costs of the balancing services with a dual imbalance pricing scheme. used by the system operator. This is the approach When dual imbalance pricing is applied, followed in the United States23 and some EU sys- portfolios to conform to a BRP gives a tems (e.g., Germany). With this mechanism, the competitive advantage to large companies and price that the BRP pays when it is short and the introduces entry barriers to small providers. price the BRP receives when it is long are the same for each point in time and space, and it is determined by the price of the balancing ser- Balancing responsibility vices. This way, settling imbalances results in a The imbalance responsibility and the imbalance zero-sum operation for the system operator. settlement are two of the most important and Sometimes, it is pointed out that the single im- controversial cornerstones of balancing mecha- balance pricing scheme can lead to speculative nism design. behaviour where BRPs may seek to deviate in real Imbalances are measured and settled at different time to be remunerated for helping the system. aggregation levels depending on the system. In System operators fear this potential behaviour

23. This is true, at least, for small imbalances. For large imbalances, depending on the system, there are additional, explicit penalties.

68 WHOLESALE MARKET DESIGN

because, although it may help balance the system, A dual imbalance pricing penalises imbalanc- it can also result in a situation where the TSO has to es, but a generation company can avoid imbal- balance two imbalance origins: the original imbalance charges by compensating the (positive or ance and the “speculation imbalance” if the BRP’s negative) imbalance with another plant within imbalance estimation is wrong (SWECO, 2015). its BRP. 24 This gives a competitive advantage to large companies in comparison to smaller ones, The dual imbalance pricing scheme seeks to and can effectively create a barrier to small DER avoid the previous problem by reinforcing the in- providers and aggregators. centives to prevent deviations from the gate closure programme. This scheme applies an addi- If dual imbalance pricing cannot be avoided, the tional penalty, on top of the price representing BRP should be defined on a unit-by-unit basis, so the balancing procurement costs, if the BRP de- as not to create these competitive disadvantages. viation is opposite to grid needs. For example, if The reference market to measure imbalances the BRP is short on production when there has been a negative system imbalance in real time, Imbalances in the real-time operation are usually the BRP will not only pay the costs of the balanc- measured with respect to the programme coming services used by the system operator, but also mitted at the gate closure (roughly speaking, the a penalty. This pricing scheme avoids speculative last intraday opportunity in Europe and the re- behaviours, for, on average, the losses when pe- al-time market in the United States). nalised will not compensate the profits when the There may be, however, additional penalties for imbalance helps the system. This scheme also deviating in real time from the programmes de- provides incentives, beyond the true short-term clared in previous market sessions, for example, costs, to accurately forecast intermittent produc- for deviating from the day-ahead programmes. tion at the gate closure. Using other markets can help allocate some sys- There is a growing consensus that if the imbal- tem costs. Operating reserves are often procured ance price does not exactly reflect imbalance precisely to deal with potential deviations from costs, this distorts the real-time signal to agents the day-ahead market. This is the approach of the and, therefore, may not provide correct incentives cost allocation design implemented in some US to encourage flexibility (Olmos et al., 2015). That systems, such as the Pennsylvania Jersey Mary- being said, if speculative behaviour is seen as a land Interconnection (PJM). Such cost allocation real threat, there is a second-worst alternative to provides additional incentives to provide accu- dual imbalance pricing, which is not allowing par- rate programmes in the day-ahead time frame, ties to speculate under strict regulatory supervi- although it makes intermittent resources less sion. Such an approach is not recommended if it competitive. ends up artificially limiting market agents’ range Balancing responsibility for RES of action. Traditionally, variable RES generation has been Defining balancing responsible parties when exempted from balancing responsibility in many coupled with dual imbalance pricing countries. The main reason for exempting RES Defining BRPs is a contentious issue when cou- from this responsibility, and socialising these pled with a dual imbalance pricing scheme (if a costs, has been to encourage RES generation by single imbalance pricing is in place, the level of further reducing investor risk. aggregation allowed does not interfere with mar- In the United States, the FERC issued Order ket signals). 890 (FERC, 2007) that, among other things, re- vamped the energy imbalance provisions in Or- der 888. Order 890 basically provides a progressive scheme by which larger imbalances pay

24. However, depending on the mix available in the BRP, paying the imbalance charges may be more economical than self- compensating the imbalance.

69 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

larger imbalance charges, but also establishes and consumption, and dual pricing for conven- softer conditions to protect RES.25 tional generation. 2 Reducing investor risk at the expense of distort- 2.3.2. Balancing products ed shortterm signals was considered an acceptable trade-off when RES volumes were small. But as volumes of variable RES generation increase, When designing balancing products, these aspects may become significantly hard to it is helpful to: control. For this reason, there is a growing con- · Define innovative products to unlock new sensus on the need to make RES increasingly flexible resource potential, responsible for their imbalances. This is an im- · Give different price signals to resources portant step towards the greater integration of performing differently, variable RES. · Separate the procurement of, balancing Despite this global trend, there are still some coun- energy, upward reserves and tries that do not expose RES producers to balanc- downward reserves, ing responsibility (See Figure 2.21 on EU countries). · To the extent possible, avoid limiting participation based on size or technology Figure 2.21 RES balancing responsibility

Significant barriers to the participation of DER Balancing responsibility for RES and RES in balancing markets can be found in the details of product definition. To properly under- Country Balancing responsibility stand this issue, it is helpful to review how bal- Belgium Yes ancing products are defined. Three criteria are Denmark Yes Croatia No relevant: 1) whether it is a capacity- or an ener- France No gy-based product; 2) whether it involves an up- Germany FIP Only ward or downward balancing capacity/energy or Ireland Partly not and 3) the time parameters involved. Italy Partly Balancing capacity and balancing energy Poland Yes Balancing capacity gives TSOs the possibility of Portugal Yes activating a certain amount of balancing energy in Spain Yes real time. This refers only to capacity reserved in Sweden Yes advance for its subsequent use in real time.

Source: European Commission, 2016b Balancing energy, on the other hand, refers to the actual variation of generation/consumption used In several market designs, the calculation and in real time to fix imbalances. pricing of imbalances may differ among conven- Upward and downward balancing products tional generation, consumption and renewable generation. Some countries, such as Italy, have Upward balancing products are procured to com- opted for applying single pricing for variable RES pensate a lack of generation in real time (or an excess of consumption).

25 FERC Order 890 requires that imbalances of less than or equal to 1.5% of scheduled energy (or up to 2 MW) be netted monthly and settled at the transmission provider’s incremental or decremental cost. Imbalances between 1.5% and 7.5% of scheduled energy (or between 2 MW and 10 MW, whichever is larger) are settled at 90% of decremental costs and 110% of incremental costs, respectively. Imbalances greater than 7.5% (or 10 MW, whichever is larger) would be settled at 75% of the system’s decremental cost for overscheduling imbalances or 125% of the incremental cost for underscheduling imbalances. Intermittent resources, however, would be settled at 90% of decremental costs and 110% of incremental costs for imbalances greater than 7.5% or 10 MW (PJM, 2013). Apart from the previous regional applying regulation, additional RES exemptions can be implemented in each particular system. For example, under NYISO’s market rules, if a variable generation resource is scheduled a day ahead, the resource must buy or sell deviations at real-time locational marginal prices. However, for up to 3 300 MW of installed wind and solar capacity, this generation is exempt from undergeneration penalties when its output differs from that scheduled in real time during uncon- strained operations.

70 WHOLESALE MARKET DESIGN

Downward balancing products are products pro- Box 2.15 shows some time parameters that are cured to compensate an excess of generation used to define the standard characteristics of the in real time (or a lower-than-programmed con- different products in the EU context (in the Unit- sumption). ed States the parameters are similar).

Type of response (time parameters) Once the previous characteristics have been dis- In order to deal with disturbances, the system op- cussed, we focus on considerations to be taken erator usually makes use of three types of bal- into account in the design of balancing products ancing products in a sequential process based on to ease the integration of RES and DER (Olmos successive layers of control (ENTSO-E, 2014). The et al., 2015; NREL, 2016). Subsections that follow names of these products vary by region: draw recommendations for such design.

•• Primary (US) or frequency containment reserve Innovative products to unlock flexible resources (EU) The definition of new reserve products is a major •• Secondary (US) or frequency restoration re- discussion in the United States and in the Euro- serve (EU) pean Union, both to unlock existing flexibility and •• Tertiary (US) or replacement reserve (EU) also to allow new flexible resources to efficient-

Box 2.15 Defining the characteristics of balancing products: The European approach

Transmission system operators (TSOs) have to define at quested power of the TSO. It is expressed in seconds if least the following standard parameters in accordance the bid is not divisible and expressed in megawatts per with the Network Codes on Electricity Balancing and second (MW/s) if the bid is divisible. Load Frequency Control and Reserves (LFC-R) to allow Full delivery period: the sum of [3] ramping period; [5] minimum and maximum duration of delivery period and [6] deactivation period. Figure 2.22 Standard characteristics of balancing products −− [3] Ramping period: (as described above). −− [5] Minimum and maximum duration of delivery period: Quantity the time during which the balance service provider delivers the full requested power to the system.

[6] Deactivation period: the time from the start of phys- 4 −− ical deactivation of the unit until the full instruction MW has been delivered; expressed in seconds if the bid is Time 1 not divisible and MW/s if the bid is divisible. 2 3 5 6 Minimal duration between the end of deactivation period and the following activation, which allows time to recover Full Activation time (from LFC-R) the capacity to provide the service once again.

Other characteristics are as follows.

Full Delivery Period Divisibility: The minimum divisible unit of balancing energy expressed in MW for the divisibility of volume and the exchange of balancing products (see Figure 2.22). expressed in seconds for the divisibility of the delivery period. Full activation time: The sum of [2] preparation period and [3] ramping period. Validity period: The period defined by a beginning time (hh:mm) and an ending time (hh:mm), when the bid could [2] Preparation period: time required prior to delivery of −− be activated. The validity period is at least the full deliv- the first megawatt (MW). ery period. [3] Ramping period: time when the bid is physically ac- −− Mode of activation: Manual or automatic. tivated, delivers the first MW and approaches the re- Source: ENTSO-E, 2014

71 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

ly participate in electricity markets. The penetra- provide more efficient signals and result in an tion of RES technologies increases the need for overall more economic system operation. The ex- 2 fast-ramping reserves, so this discussion is sup- perience of defining new products for which bat- posed to become more important in the future. teries can provide fast-responding reserves has For example, pioneered by CAISO (Galiteva and been successful in PJM (see Box 2.16). Casey, 2015), there is a growing trend in the Unit- Different price signals ed States to consider introducing requirements In many US systems, it was usual to have differ- for “flexibility reserves”. As pointed out, these ent resources following regulation signals with flexible reserves encompass a family of products varying performance, but all equivalently clear- that do not (yet) have a universal definition. ing on the market under the same remuneration In CAISO and MISO, flexibility reserve products (Benner, 2015). This was detected as a major con- are typically designed to better prepare the sys- straint on new resources capable of providing tem for expected and unexpected ramping, not better responses since that extra value was not to resolve energy or capacity shortfalls. acknowledged on the market. FERC Order 755 (FERC, 2011), a performance-based regulation, Some simulations (Krad et al., 2015) suggest that changed this by measuring performance with a these products will keep more resources available standard metric and ranking the clearing based in advance of ramping events (see Figure 2.23, on performance and benefits. which shows how the available fast-ramping capacity increases when flexibility products are The implementation of the performance-based traded) and likely reduce scarcity price spikes as- regulation has been slightly different from one sociated with ramping shortfalls. ISO to another, but in general, it has removed discrimination against flexible resources, enabled energy storage resources to profitably provide Figure 2.23 Available ramping capacity with regulation services, and reduced regulation re- and without flexibility reserve products serve costs (Xu et al., 2016).

Separation of balancing capacity and balancing 7 energy products 6 When balancing capacity and balancing energy 5 products are procured jointly, only the balancing 4 service provider having sold balancing capacity 3 can provide balancing energy in real time to the 2 system operator. Examples of markets with this

Ramp [MW/5 min] Ramp [MW/5 1 design include the Spanish and the Danish auto- o matic frequency restoration reserve (FRR) markets, and the German automatic and manual FRR 106.4 106.6 106.8 107 107.2 107.4 markets. Time [h] The joint procurement of balancing capacity With ﬂex Without ﬂex and balancing energy products is often ineffi-

Source: Adapted from Krad et al., 2015 cient because it does not reveal the benefits of the most-cost-efficient resources in real time. But New, flexible resources may need new product what is even more important is that it limits the definitions. This is the case, for instance, with bat- potential participation of renewable producers teries, which can offer a quick response but can- and other DER participants. not retain the response for long periods of time. In general, the procurement of capacity products A new product designed to unlock battery poten- is carried out well before real time, so the system tial can be compatible with current mechanisms, operator can ensure in advance the availability of

72 WHOLESALE MARKET DESIGN

Box 2.16 Defining innovative products for batteries in the United States

Following FERC Order No. 755 (FERC, 2011), the Pennsyl- vania Jersey Maryland Interconnection (PJM) implement- Figure 2.24 Conventional regulation signal (blue) ed a performance-based regulation. Resources providing and the faster regulation D signal (yellow) frequency regulation are remunerated based on their performance – how fast and accurately they respond to PJM’s signals. Faster and more accurate resources receive higher compensation.

PJM now has two separate signals. Conventional generation resources follow a traditional slower signal (Reg A). Faster resources, such as batteries, follow a dynamic or fast-responding signal (Reg D). 1 Hour

This separate fast regulation signal (Reg D) was de- Conventional regulation signal Faster regulation-D signed to provide zero net energy required over a time Source: PJM, 2013 period shorter than one hour (see Figure 2.24). Thus, this product is energy neutral, meaning that the accumulat- tice, this calls for defining a rate of substitution between ed signal will converge back to zero after a short time. traditional Reg A resources and dynamic Reg D resources The implementation of this new signal removes a serious or, in other words, comparing the value of fast moving re- obstacle to the participation of energy storage in PJM’s source‘s regulation with that of traditional reserves. regulation market. This has been carried out by means of the so-called ben- This new market design uses two separate payments – efits factor (BF) curve (Figure 2.25). This curve expresses one for capacity and one for performance. As of April the equivalence of both types of reserves and depends on 2015, there are several energy storage resources par- the amount of fast reserves procured. The x-axis represents ticipating in PJM’s regulation market. Regulation D was the percentage of total reserves covered through fast-regu- developed specifically for energy storage devices with lation reserves, while the y-axis represents the relative value limited storage capabilities with respect to the maximum of these fast reserves with respect to conventional reserves. capacity. (For example, lithium ion batteries typically op- As shown in the BF curve, the first MW of fast reserves (0% erate to a 4:1 MW:MWh ratio.) That said, other technolo- in the x-axis) is worth 3 MW of slow reserves (y-axis). How- gies can also provide such services (since 2014, hydro- ever, if the amount of fast reserves increases, this equiva- electricity can qualify for Regulation D). lence lowers (even down to zero, for 40%).

The major challenge associated with defining different As also shown in the figure 2.25, the BF curve had to be products updated recently because the equivalence acknowledged during its early implementation proved to be too gener- The major problem with this design is how to define the re- ous and was leading to suboptimal results. quirements for both types of regulation services. In prac-

Figure 2.25 Updated benefits factor curve to optimise operations under all system conditions

Updated Benefits 3 Factor Curve 2.5 New Benefits Factor Curve 2 Current Benefits Factor Curve 1.5 Benefits Factor Cap for ‘Excursion’ Hours (26.2%) 1

Resources will be New Reg-D Cap (40%) 0.5 assigned unit speciﬁc BF down the BFcurve, Performance Adj. MWs as today 0 (o er MW* Preformance Score) - 0% 7% 14% 21% 29% 36% 43% 50% 57% 64% 71% MW and % of Regulation Requirement (700MWs) 0 50 100 150 200 250 300 350 400 450 500

Source: Adapted from PJM, 2013

73 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

sufficient resources to deal with real-time imbal- It is also an inefficient design, given that pow- ances. For instance, in Spain, balancing capacity er systems, especially those with significant RES 2 is procured one day before real time; in Germa- penetration, can have very different require- ny and Austria, balancing capacity is guaranteed ments for upwards and downwards reserves in a one week before real time. In the case of intermit- given period. tent renewable generators, committing capacity Bid size requirements within those time frames is a clear barrier that can hamper their participation for all intents and A minimum bid size is often imposed for partic- purposes (they cannot forecast their production ipation in balancing mechanisms. Such a thresh- accurately in such a time frame, as seen in Sec- old is commonly introduced for the sake of sim- tion 2.2 of this chapter). plicity and in order to maintain the computational effort of clearing the market at an acceptable A more efficient option is to decouple capac- level. However, depending on the minimum size ity and energy provision. It is more efficient to requirement, smaller resource providers may be procure, in advance, exclusive capacity from dis- prevented from participating. This is particular- patchable units (and to ensure that those bal- ly true if, concurrently, the aggregation of indi- ancing service providers providing capacity are vidual units’ offers (to comply in aggregate with obligated to offer the amount sold on the balanc- minimum bid size) is not allowed. Even if a re- ing energy market, perhaps by including a price duction of the size requirement would make the threshold for them). At the same time, all resourc- market clearing more complex, it may effective- es, including those with no balancing capacity ly increase the participation of RES and demand committed, may participate in the balancing en- resources. ergy market. This way, a variable RES unit, which cannot sell balancing capacity much before real In Germany, the minimum bid size is 1 MW; in Bel- time without incurring a high risk, would be able gium, 4 MW; and in the the Kingdom of Nether- to sell surplus energy production (above fore- lands, 5 MW. The highest minimum bid size is that casts) as balancing energy. implemented in Spain, which is 10 MW. Examples of this design include the Belgian and Technology-specific products the Dutch automatic and manual FRR markets The only prerequisites to providing a product and the Danish manual FRR market. should involve objective technical evaluations and tests. Allowing only certain technologies to Separation of upwards and downwards reserves provide a product, or similarly, not allowing cer- In some systems, the reserve capacity product tain technologies to offer one (as is the case for links the amount of upwards and downwards re- RES and DER on many markets) reduces compe- serve capacity that has to be provided in what tition and the efficiency of the overall dispatch. is sometimes called the “regulation band”. The Spanish, Italian and the Danish system operators By the same token, the aggregation of resourc- procure this band reserve product for the FRR es should be allowed to participate as long as balancing capacity. compliance with the objective requirements is demonstrated. The definition of a single product for providing upwards and downwards reserve imposes clear barriers to the participation of renewable generators since the (opportunity) costs incurred by RES for providing upwards reserves are much higher than those for providing downwards reserves.

74 WHOLESALE MARKET DESIGN

2.3.3. Reserve markets and pricing then the solution should not be to change the of the products pricing rule but rather to fix the situation (or return to a regulated provision of the service). When sufficient competition can be ensured, Marginal prices may be more volatile, but they are and when products are homogeneous, the correct signals that reflect the real-time value balancing services should be priced based on of flexibility and give the right incentives to the pay-as-cleared schemes. resources that can respond to these price signals. If market power is a problem, relying on a Furthermore, marginal prices provide efficient market mechanism might not be enough. long-term incentives.

As was analysed in Section 2.2.5, most Europe- 2.4. LONG-TERM SUPPORT an systems do not procure energy and reserves MECHANISMS in a fully co-optimised way. In the United States, Long-term mechanisms refer to those regulatory meanwhile, the procurement of these products is instruments whose purpose is to guide genera- usually co-optimised. The efficiency of one or the tion expansion according to the strategic view of other process has already been analysed. governments and regulators. This strategic view In some cases, the provision of a service (typical- includes capacity (or adequacy) mechanisms and ly primary regulation) is obligatory and not remu- support mechanisms for RES. nerated, while secondary regulation capacity and In general, generation capacity mechanisms and additional upward regulation capacity are pro- some RES support mechanisms provide investors cured through market mechanisms. with an additional and/or more stable remunera- The pricing of balancing products can either be tion. However, there has been a major difference based on pay-as-cleared or pay-as-bid mecha- between them. Conventional generation tech- nisms. In many cases, the procurement is bilateral nologies could be financially viable in a well-de- and so, by definition, these products are pay-assigned and fully functional energy-only market, bid. In other cases, pay-as-bid has been imple- i.e., they would be able to recover their total pro- mented (and average reserve pricing) with the duction costs, including a suitable rate of return 27 objective of mitigating market power and reduc- on investment. On the other hand, in the ab- ing the volatility of prices. Moreover, pay-as-bid sence of a proper internalisation of environmental clearing rules are mandatory when the products/ externalities, until very recently, some RES have services procured by the TSO in a specific auc- not been competitive enough to be financially vi- tion are not adequately homogeneous in terms able in electricity markets, either energy only or of quality, location and time. Indeed, the adop- with generation adequacy instruments. As a re- tion of a marginal pricing rule requires that the sult, they have historically needed an additional market has been segmented into homogeneous support mechanism. Nonetheless, this does not products/services and that a relation of substitu- automatically imply that they should be banned tions among the different products has been es- from general capacity adequacy instruments, as tablished. discussed in Section 2.4.1. In general, the market price should be calculated Next, we introduce the capacity (or adequacy) applying the marginal pricing.26 If it is believed mechanisms and the support mechanisms for that the market is prone to market power abuse, RES and later review the recommended trends in their design today.

26. When it is impossible to separate the different products in groupings of similar products (because of different qualities or times of negotiation), the use of pay-as-bid rules is justified. 27. This does not necessarily mean that in the absence of a fully active demand, the levels of reliability achieved might not reach those desired by the regulator.

75 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Long-term capacity mechanisms Support mechanisms for RES A combination of market and regulatory failures Many of the most promising forms of RES once 2 is often believed to threaten the long-term se- faced significant barriers to growth, ranging from curity of supply. If this is the case, to the extent their high capital cost and perceived risks to a possible, the failures should be identified and re- market and regulatory structure designed to ac- moved to allow the energy-only market to work commodate conventional fossil-based genera- and give proper long-term incentives. From an tors. Policy makers to date have recognised that investment perspective, long-term markets and in order to boost their deployment and eventually long-term price signals are as important as prop- compete on the energy market, RES needed spe- erly functioning short- and very-short-term mar- cific support policies. kets (European Commission, 2015b).28 Well-func- Regardless of the mechanism to which RES re- tioning long-term markets are a prerequisite for sources finally have access, it is important that a fully functional energy market.29 Long-term renewable technologies are exposed to market hedging products are relevant for all market par- signals as much as possible in order to achieve ticipants and particularly essential for indepen- their actual integration in power systems. They dent non-vertically integrated agents. can still receive economic support if needed, but A liquid and efficient forward market helps pro- they should be increasingly required to partici- vide optimal long-term price signals about fu- pate in all the segments of the market, including ture market expectations. This, in turn, promotes the short and long term, and thus be responsive long-term security of supply in a market-compat- to all market signals. ible manner. Despite this essential role, well-func- The ideal RES support framework should allow tioning long-term markets do not happen sponta- and foster this participation in the market, and neously in electricity markets, and this represents should take into consideration the revenues ob- one of the major market failures affecting the tained by renewable plants in all the markets (in- long-term security of supply. cluding energy, capacity and other services), in In comparison with an energy-only framework, order to calculate the incentive required to make long-term support mechanisms can help in re- the investment attractive and to make it available ducing the investment risk premium. In the ab- to the project developer. sence of well-functioning long-term markets, a There are two main aspects to consider here: 1) capacity remuneration mechanism can, in some as far as technically possible, and taking into con- cases, result in lower, but more stable, income for sideration their specific characteristics, mature generators. RES technologies should increasingly participate In this section, we analyse how the regulatory in all electricity markets and be exposed to the design of these mechanisms should evolve and same incentive mechanisms as any other gener- adapt to the future scenario, in which high pene- ation technologies (including capacity markets) tration levels of renewables and DER are expect- and 2) if RES need technology-oriented support, ed. It is important to bear in mind the experience the corresponding long-term mechanism should accumulated over the past two decades in the be designed so that any distortion of the electric- regulation of market mechanisms for generation ity markets is minimised. supply and RES support instruments. Today, the major objective of regulatory design must be to improve the performance of these support mechanisms while minimising their impact on the correct market signals.

28. We refer here to time horizons between one and 15 years. 29. This involves long-term energy markets and also long-term carbon markets (e.g., the role of a properly designed emission trading scheme is fundamental in the EU context).

76 WHOLESALE MARKET DESIGN

2.4.1. Generation adequacy of variable renewable technologies (and, in some mechanisms and RES integration cases, generation overcapacity), hampers investments as much as uncertainty regarding the long-term deployment of these resources (due RES technologies should be allowed to regulatory uncertainty as regards long-term to participate in generation adequacy planning and to uncertainty of learning curves). mechanisms and be exposed to their market This is certainly the case in those systems where signals. RES support policies have been unpredictable because they did not stick to any consistent en- The penetration of renewable energy technolo- ergy policy plan. gies in power systems is often argued to be one On the other hand, renewable technologies also of the key reasons for the implementation of a represent a valuable resource for security of sup- generation adequacy mechanism. This is espe- ply (Gouveia et al., 2014): cially the case in Europe, where this penetration has already reached significant shares. RES, •• Regional markets can help manage fluctuations whose installation has been driven, so far, by pol- in their electricity production by increasing icy support, are claimed to depress prices in the the geographical scope and thus reducing the short-term electricity market (Moreno et al., 2012; probability of a concurrent lack of renewable Reuter et al., 2012). Furthermore, RES technol- energy. Developments in storage technologies ogies have caused a significant decrease in the can also reduce these fluctuations on a smaller load factors of many conventional plants, as can scale for each unit or plant. be observed in Figure 2.26. These combined ef- •• Renewable sources may complement other en- fects, together with their increased short-term ergy resources, and RE technologies construc- volatility, are often claimed as major reasons be- tion time may be significantly shorter than that hind the need for capacity mechanisms. However, of other technologies, as previously discussed this is not completely correct. (Barroso and Batlle, 2011). • The cost of RES technologies, especially that Increased price volatility in the short term, or the • of wind and solar PV, is swiftly decreasing, and reduction of average prices due to the installation

Figure 2.26 Annual gross aggregated electricity production from gas-fired, solar and wind electricity plants in the European Union

GWh 700,000 600,000 500,000 400,000 300,000 200,000 100,000 0 1991 2011 1997 1992 1993 1995 2012 2013 2015 1998 1996 1999 1994 2014 1990 2001 2010 2007 2002 2003 2005 2008 2006 2009 2004 2000

Natural Gas Wind Solar

Source: Adapted from ACER-CEER, 2016b

77 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

these technologies are already, or will soon be, short-term market, as the remuneration is thought able to compete with conventional technolo- to complement wholesale market revenues.30 gies in several electricity markets. 2 In this context, the short-term variability of some Depending on the type of technology and the de- renewables does not impede them from con- sign of the mechanism, renewable technologies tributing to system reliability. Obviously, once could also be able to participate in generation they are installed, their non-dispatchable nature adequacy mechanisms, thus also competing with does not leave much room to manage them in conventional technologies in the capacity market. the short run, but if proper incentives are imple- Nonetheless, the ability of some RES technolo- mented, the investment process can be oriented gies – like wind and solar PV – to contribute to to maximise system reliability (for example, by the security of supply and to help relieve scarcity opting for those RES technologies or those sites conditions is generally lower per unit of installed whose resources are positively correlated with capacity than for conventional technologies. The the net needs of the system). If they are able to RES contribution depends on the specific condi- deliver, on average, their expected contribution in tions of the power system where they are locat- the medium term, they permit the saving of water ed and on the typology of scarcity conditions, as in the reservoirs, regardless of the daily or hourly analysed next. schedule of their production. Their participation in a generation adequacy mechanism can there- Variable generation in energy-constrained fore be beneficial (see Box 2.17). and capacity-constrained systems In energy-constrained systems, e.g., in hydro- In capacity-constrained systems, scarcity condi- power-dominated countries, the coverage of the tions arise because there is not enough installed instantaneous peak load is not an issue, and scar- capacity available to meet the load at a given city conditions are more related to dry seasons moment. Aggregating all the hours, the system that could last for months. The system could cer- could certainly have enough energy available tainly satisfy peak demand, but it would be un- to satisfy demand on a given day (more than able to supply demand during the remaining enough prouction capacity in off-peak hours), hours of the day/week. In this context, rationing but it lacks installed and available capacity to sat- takes place due to a lack of available energy, not isfy peak demand. This type of potential scarcity capacity. This situation is quite common, for ex- condition is found in many European and North ample, in Latin America. Brazil is the paradig- American power systems and has prompted op- matic example, with its huge hydropower facili- erators of these systems and market participants ties and multiyear storage capability. It must be to model the very-short-term time frame in great kept in mind, then, that in an energy-constrained detail. Generation adequacy mechanisms are also system, the reliability product procured in a ca- focused on this time frame, which can be easi- pacity mechanism is not capacity. For example, ly observed in the reliability product design. Es- in some power systems in Latin America (as in pecially in the most recently introduced capacity Brazil, Chile or Peru), generation adequacy mech- remuneration mechanisms (CMs31) – for example, anisms take the form of long-term electricity auc- the British capacity market or the pay-for-perfor- tions in which the reliability product is energy, or mance reform of the Forward Capacity Market a call option on future energy (as in Colombia). of ISO-NE) – the reliability provider is required In some cases, these auctions somehow “sub- to deliver its contribution during stress events. stitute” for the energy market (as investors are These may be identified by the system operator fully hedged from the fluctuations of short-term either at short notice (e.g., four hours in advance market prices); in other cases (e.g., Colombia), in the United Kingdom) or no notice at all (the the auctions let agents trade electricity on the case of the so-called “shortage events” in the Forward Capacity Market).

30. See Maurer and Barroso (2011) for details. 31. Capacity remuneration mechanism is the term used in the EU context to refer to long-term security of supply mechanisms. In the US context, the term resource adequacy mechanism is used instead.

78 WHOLESALE MARKET DESIGN

Figure 2.27 Coincidence of solar energy with peak Box 2.17 Wind penetration in Brazilian demand in Colorado long-term auctions

Brazil is a typical example of an energy-constrained 14,000 700 system. Its market is organised around different 12,000 600 kinds of long-term electricity auctions, which de- 10,000 500 termine the majority of the economic flows, leaving 8,000 400 only minor settlements to the short-term market. 6,000 300 These tenders also represent the generation ade- 4,000 200 Net Load (MW) Load Net quacy mechanism used in this country. 2,000 100 Solar Generation (MW) Solar Generation 0 0 The outstanding participation of wind energy in the 1 12 24 36 48 60 72 Brazilian auctions has been celebrated worldwide Base case net load Net load with PV PV for the low prices offered by this technology. These prices are related to the very high capacity factors Source: Adapted from NREL, 2013b considered in the project planning, which range from 30% to 60%. This high capacity factor not only describes the amount of energy that each unit can This information can be used to define de-rating sell in the long-term auction, but it also represents factors that constrain the amount of capacity that its expected contribution to generation adequacy. RES can trade in a capacity mechanism, thus lim- In fact, in the scarcity conditions foreseen in Brazil, iting the remuneration they can receive from it. the variability of wind power does not impede these plants’ contribution to system reliability. A well-designed regulatory framework for RES development should be able to allow renewable Also, Brazil’s penalties for underperformance reflect participation in CMs in capacity-constrained sys- the need for guaranteeing that enough energy is tems, leaving to investors the definition of their available during the year, more than that required to risk-hedging strategy and the decision of wheth- cover the peak load. The penalties are also for delays er to enter into CM contracts or not (see Box 2.18). in the construction of plants. Once installed, a minor settlement is carried out every year, but the main Generation adequacy and interactions with RES penalty scheme is represented by a cumulative four- support mechanisms year performance assessment (IRENA, 2015), which As has been mentioned, the debate over wheth- clearly reflects the time dimension of the generation er or not to isolate RES from market signals (and adequacy problem in Brazil. to what extent) has been intense.32 The inefficiencies derived from reduced wholesale market exposure were assumed to be outweighed by In this context, the variability of renewable pro- a reduction of investors’ risks. However, as RES duction is an issue because if the RES are not penetration has increased, the maturity of RES available at the very moment when the system technologies has reached a level that allows these stress occurs (e.g., due to a lack of wind or so- plants to be largely exposed to market mecha- lar radiation), then they cannot contribute to re- nisms. At the same time, the impacts of a lack of ducing the shortage. The contribution of variable market integration on the broader power system RES to capacity mechanisms in capacity con- have become more difficult to ignore.33 strained systems has to be evaluated using specific, well-known and widely available methods In principle, CMs should allow RES of all types to (for solar and wind capacity values, see NREL participate in any kind of generation adequacy (2013b; 2008); Figure 2.27). Statistical analyses mechanisms. This is already taking place in Bra- can be used to calculate the likely production zil, where generation adequacy mechanisms are from RES during the expected scarcity periods. a component of the income of RES. RES technol-

32. For example, in Europe the debate was around whether to use feed-in tariffs, tradable green certificates or auction-based mechanisms – see, for instance, Butler and Neuhoff (2008). 33. See Fraunhofer ISI (2015) for a discussion of the growing impact of negative prices on European wholesale power markets.

79 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 2.18 Renewable participation in PJM’s reliability pricing model 2 The Pennsylvania Jersey Maryland Interconnection (PJM) qualification phase prior to the auction. These factors, is a large interconnected (capacity-constrained) pow- lower than those in the Brazilian case, represent the ex- er system in the United States, whose generation mix is pected contribution of these units during scarcity con- mainly based on conventional thermal plants. Its gener- ditions in the PJM. Since these conditions are related to ation adequacy needs are covered through a capacity short-term shortages, the likelihood of RES resources be- market called the reliability pricing model (RPM). RPM ca- ing available at the specific time when scarcity occurs is pacity providers, selected through centralised auctions, estimated to be low. commit to be available whenever the system operator As for penalties, a capacity remuneration mechanism requires their contribution to solve scarcity conditions. reform being implemented in the PJM (as well as in the In the 2017/18 RPM base residual auction, the main pro- ISO New England), inspired by the “pay-for-performance” curement process of this mechanism, 803 MW wind pow- principle (see Mastropietro et al., 2015a, for details), in- er plants and 116 MW solar resources were cleared (PJM, troduces severe penalties for underperformance. It is not 2014). On average, wind and solar units were assigned clear how this will affect renewable participation in future a 13% and 38% capacity factor, respectively, during the auctions.

ogies started being cleared in specific renewable electricity rationing. This side of the mechanism auctions, but they also ended up being selected is now being reinforced through the implemen- in conventional long-term electricity auctions for tation of performance incentives. If the capaci- new energy. As analysed in the next subsection, ty mechanism design is robust and the reliability this convergence is not yet complete, and some product is technology neutral, RES participation rules still need to be harmonised. in system reliability would be acknowledged – and RES would be exposed to an economic sig- Meanwhile, in capacity-constrained systems, re- nal that prompts them to be available when the newable energy technologies have so far partic- system most needs them. If specific renewable ipated in CMs to only a limited extent. In many energy technologies/sources are technically able countries, regulators decided that RES, which to compete with other resources on a level play- were already receiving some other kind of incen- ing field, relevant renewable power plants should tive (grants, feed-in tariffs [FiTs], etc.), were not not be favoured, either through their exemption eligible for the capacity mechanism remuneration. from penalties or through overgenerous recog- This was the case, for example, in the UK’s Capaci- nition of firm energy/capacity. If RES units have ty Market (DECC, 2014). The rationale for this deci- to compete with conventional technologies, they sion is that the capacity mechanism remuneration should be subject to the same rules. RES support represents an investment incentive. Renewable mechanisms are not incompatible with this ap- technologies, for which investment is already be- proach. As mentioned above, the ideal RES sup- ing incentivised by some sort of RES support port scheme would deduct the estimated reve- mechanism, are therefore not eligible since they nues obtained by the renewable plant on all the do not need further investment support. markets, including the energy and capacity mar- This argument misses the main point that we are kets, from their total costs. This would be done to making here – that RES technologies need to be calculate the incentive required to make the in- exposed to market signals. They should be investment attractive, and to make it available to tegrated into the wholesale market as much as the project developer, without spending unnec- possible. Capacity mechanisms provide not only essary funds. an investment incentive, but they are supposed However, in real markets, several factors can to provide market agents with incentives to fos- complicate this simple theoretical discussion. ter the availability of committed resources during According to some stakeholders, if “subsidised” scarcity conditions, with the objective of avoiding

80 WHOLESALE MARKET DESIGN

Box 2.19 Renewable participation in ISO New England’s Forward Capacity Market

ISO-New England pursues generation adequacy through “protect” the capacity market from excessive renewable its Forward Capacity Market. This mechanism has been participation, which could depress capacity prices, the recently reformed to introduce stronger penalties for un- regulator introduced specific offer review trigger prices derperformance (FERC, 2014d). for renewable technologies, which apply to all renewable bids exceeding 200 MW in each auction. These trigger Renewable penetration may grow significantly in the next prices are intended to prevent uneconomic or subsidised few years (wind represents 42% of generation project new entries from distorting market prices by setting a proposals). According to ISO-NE (2015), state subsidies price floor below which new entrants must demonstrate for renewable resources will put downward pressure on their costs or withdraw from the capacity auction. energy-market prices, but the capacity market will help balance the revenue needs for conventional resources. To

renewable technologies are allowed to take provision can heavily impact the overall effective- part in the capacity market, it may be difficult ness of the scheme in achieving its objectives.34 to guarantee that they do not depress prices in If renewable and conventional technologies have capacity auctions (as they have already done in to compete for the provision of generation ade- the day-ahead market in some power systems). quacy, they have to be at the same level. When This concern is further described in Box 2.19, possible, no discrimination should exist in the which focuses on ISO-NE’s experience. definition of the reliability product, and the same The justification for the floor described in Box commitment should be required from all reliabili- 2.19 is flawed. The objective of a capacity mecha- ty providers, regardless of technology. nism is not to compensate all generators for their Two design elements must be carefully consid- investment cost, no matter the resulting mix and ered in order to avoid discrimination between re- reserve margin in the system. Instead, the aim is newable and conventional technologies. The first to provide investors with incentives – when need- is the performance incentive to which reliability ed – to ensure that a capacity level that maxi- providers are exposed, i.e., the penalty for un- mises the social benefit is installed. If there is an derperformance. The second is the methodology excess of capacity due to wrong investment de- used to calculate the firm energy/capacity that cisions, generators will see the prices depressed resources are allowed to trade in the generation in the short and medium term. In such a context, adequacy mechanism. Without a proper harmon- they will most likely ask for a higher capacity re- isation of these contract provisions, competition muneration to ensure total cost-recovery. But between renewable and conventional technolo- note that, in this case, the remuneration should gies will be distorted. This could seriously impact be close to zero since no additional incentives are the effectiveness of the generation adequacy needed to ensure reliability. mechanism in attracting investment and in guar- Rules to be harmonised in capacity mechanisms anteeing reliable electricity supply.

When signing a contract in the framework of a Regional market context generation adequacy mechanism, power resourc- The importance of exploiting the synergies be- es are committed to providing a certain reliability tween RES development and regional market in- product, which is defined in detail in the contract tegration has already been underlined. Increas- provisions. The design of the reliability product ing the geographical size of the market helps represents a cornerstone for the generation ad- manage fluctuations in RES electricity produc- equacy mechanism. Minor changes in a specific tion, reducing the probability of a concurrent

34. Batlle et al. (2015) identify the main design elements of the reliability product.

81 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

lack of renewable energy. Furthermore, regional et al. (2015b), the basic requirement is to open market integration permits a co-ordinated plan- CMs to cross-border participation. If resources 2 ning of generation infrastructures that exploits are allowed to trade their reliability in all capacity the available renewable resources as efficiently mechanisms within the regional market, regard- as possible.35 less of their point of connection (that being the only obvious constraint in the interconnection ca- A well-functioning regional market, however, pacity), they can still locate optimally. That said, should encompass all time dimensions, including some sort of co-ordination between the neigh- generation adequacy. In this respect, differenc- bouring TSOs will always be necessary (for exam- es in adequacy mechanism choices coupled with ple, to avoid one plant selling the same product in a lack of co-ordination can undermine regional different systems). market integration. International integration of national CMs is currently a major challenge (Per- This discussion is particularly relevant to the Eu- ner, 2015). ropean Union, where the regional market integration process is advanced. The EU experience is The implementation of different national CMs to achieve a range of diverse reliability targets described in Box 2.20. does not impede the optimal exploitation of regional resources but requires a minimum level of co-ordination. As mentioned by Mastropietro

Box 2.20 Cross-border participation in European Capacity Remuneration Mechanisms

In the European Union, market integration efforts have (2015b) highlighted that one main barrier to such partici- focused until now on the security and the economic effi- pation is related to mistrust in the fulfilment of article 4.3 ciency of short-term time frames, i.e., the day-ahead and of the Security of Supply Directive (2005/89/EC), which operation markets. Long-term, capacity remuneration states that “Member States shall not discriminate be- mechanisms are being introduced with little co-ordina- tween cross-border contracts and national contracts”. In tion. Their design seems to rely almost exclusively on the order to include generation from a neighbouring system domestic generation mix in the corresponding Member in a capacity mechanism, the transmission system oper- State, with the main objective of maintaining a reliable and ator of the country launching the CM must be sure that, self-sufficient power system. Obviously this is at odds with during scarcity conditions, the foreign resources are able a regional concept of the security of generation supply. to fulfil their physical supply commitment linked to the capacity mechanism. Nonetheless, most electricity laws This concern has been clearly expressed by all EU institu- and national network codes in force in the Member States tions, including in the recent consultation document from still contain clauses that maintain that exports to other the European Commission (European Commission, 2015c) countries will be interrupted in case of a domestic emer- which claims that mechanisms to ensure generation ade- gency of supply. quacy should be open to all capacity that can effectively contribute to meeting the required reliability level, includ- This barrier can be removed only through greater co-or- ing capacity from other Member States. Nonetheless, CM dination among system operators, who should commit designs implemented or proposed so far in the European to fulfilling the Security of Supply Directive through the Union (DECC, 2014; RTE, 2014; AEEG, 2011) do not allow modification of national (and regional) network codes explicit cross-border participation. Mastropietro et al. and operation procedures.

35. Obviously, this implies the existence of a regional transmission network with the adequate cross-border interconnection capacity and a regional transmission expansion strategy capable of following the evolution of the generation mix and the need for further interconnections. This relevant topic, however, exceeds the scope of this section, which is focused on electricity markets, and thus on activities open to market competition.

82 WHOLESALE MARKET DESIGN

2.4.2. RES promotion mechanisms and the two methods are similar in that they decou- wholesale market integration ple actual production from the support price and therefore do not interfere with short-term market RES technologies may continue to require signals. economic support. This support should Capacity-based mechanisms are characterised become more market compatible. Several by several key design features that affect mar- design options exist. A balance must be found ket compatibility. Four such features are outlined between optimal investment incentives below. and market compatibility. Design feature 1: Support payment, amount and timing Many of the most promising forms of RES have An obviously critical initial decision in any sup- in the past years faced significant barriers to port scheme is the overall size of the payment: growth, ranging from their high capital costs how much support does a project need to earn and perceived risks to a market and regulatory in the various segments of the market (includ- structure designed to accommodate convention- ing energy, capacity and other services)? There is al fossil-based generators. In several countries, significant risk in the decision, as an overly gen- policy makers recognised that in order to boost erous payment can lead to overinvestment in re- RES deployment with the objective of eventual- newable generation. On the other hand, too small ly competing on the energy market, RES need- an incentive may fail to attract sufficient invest- ed to be promoted with specific economic in- ment. One of the design features that affects this centives. The cost of renewable technologies has issue is the timing of the support payment, which been progressively decreasing over the past few can either be calculated ex ante or ex post. years. Soon these units may be (and in some cases already are) able to compete with convention- Most capacity-based support is determined ex al technologies without the need for economic ante based on the expected cost of a particular support. In some countries, however, incentives generation technology (as in Spain) or on the ac- for renewables are still needed while the environ- tual cost of the project (as in the United States mental externalities that RES mitigate are not in- with the Investment Tax Credit). Alternatively, the ternalised in the market prices. size of the incentive payment can be determined ex post based on actual costs and observed rev- This section discusses the three major RES sup- enues from the market. For example, if energy port scheme classifications, namely capaci- market prices turned out to be lower than ex- ty-based mechanisms, production-based mech- pected, an ex post capacity incentive could be anisms and support mechanisms for prosumers, adjusted to ensure RES plants are able to recover and their compatibility to wholesale market inte- their fixed costs. gration. Providing support upfront has the advantage of a −− Capacity-based mechanisms low administrative burden (it requires only a sin- In purely capacity- or investment-based support gle transaction) and also rewards the best-per- mechanisms, a remuneration is provided on a forming projects – the windiest/sunniest sites will per MW installed basis (i.e., there is no energy or have an easier time covering any revenue short- availability contract involved). These payments falls on the energy market and will therefore have are intended to cover the difference between a an easier time attracting investors. On the other plant’s upfront investment costs and any market hand, calculating the incentive ex post makes it revenues. While the choice of paying based on easier to guarantee a certain level of cost-recov- installed capacity or project cost has important ery, regardless of external factors, such as market consequences, depending on the policy goals, prices or plant location.

83 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Design feature 2: project relative to a reference facility. This has the Choice of reference plant effect of distributing support costs to where they 2 Closely related to the timing issue is selecting a are needed: the projects in the best (windiest) lo- reference plant for calculating the size of the ca- cations have their incentive reduced since they pacity payment. The first decision is between us- are earning above-average market revenues, ing the actual cost of a project versus using a ref- while projects in suboptimal locations receive a erence cost, derived from a “typical” plant. The premium incentive to ensure they are able to re- ITC in the United States is an example of the for- cover their investment costs (Purkus et al., 2015). mer, while support schemes in Spain, Germany Design feature 4: and Russia are examples of the latter (IFC, 2013). Minimum performance requirements Paying based on actual costs is straightforward One of the consistent criticisms of capacity-based and incurs a low administrative burden, but all mechanisms is that they do not motivate effec- else equal, this approach risks providing incentive system design, efficient component selection tives for overpriced and underperforming proj- (high-efficiency panels/turbines) or regular main- ects. The use of a reference facility helps over- tenance once the project is operational (Hoff, come these problems by creating a benchmark; 2006). These risks can be mitigated by attach- if a developer can build a plant with a lower cost ing minimum performance requirements to the or a higher performance than the reference plant, incentive and then adjusting or even withholding then it will have higher profits, thus creating a part of the payment subject to conditions (which natural incentive for higher-quality projects. could include any sort of locational signal). Developing reference plants involves estimating These tactics require ongoing performance mon- capital costs and production hours. However, re- itoring, which increases administrative costs, but lying on reference plants can be problematic if the upside can be substantial in terms of higher there are too few of them, or if the regulator fails and longer-lived production from quality systems. to properly diversify them depending on the pol- In addition to performance monitoring, or perhaps icy objectives. Spain has addressed this problem as an alternative, regulators can impose minimum by defining some 1 276 different reference plants, standards for component efficiency or system de- diversified by technology, size, location and con- sign. It can be challenging, however, for regulators struction date (Barquín, 2014). to maintain reasonable standards in light of rap- Design feature 3: idly evolving technology. Furthermore, there is a Frequency of support payment risk of regulatory overreach as innovative technol- Capacity-based support can be delivered either ogies and design approaches may be restricted in as a one-time lump-sum payment or through a their development if they do not meet the admin- series of periodic payments. One-time payments istratively determined criteria. are potentially advantageous as they can be con- Assessment of market compatibility cluded in a single interaction, reducing the ad- In general, capacity-based support mechanisms ministrative burden on both the developer and are highly compatible with a market-oriented the government counterparty. On the other hand, power generation sector. They avoid the mar- periodic payments are more flexible as total com- ket distortions of production-based schemes pensation may be adjusted over time. This can be by decoupling payment and performance. They a useful feature for controlling policy costs and also come with their own set of potential risks limiting windfall profits, or conversely, for keep- that need to be managed, such as an unstable ing an underperforming project profitable. investment environment, incentives flowing to Germany uses periodic payments to support low-quality projects and inflexibility to evolving wind generators, which are adjusted over sev- market conditions. eral years depending on the performance of the

84 WHOLESALE MARKET DESIGN

Many, though not all, of these risks can be mit- to either remunerate any and all output, or pro- igated through incorporation of specific design vide support only up to a certain limit. For ex- features. The risk of under- or overpaying can ample, premiums for wind production could be be reduced by using reference plants, diversified restricted to a benchmark number of hours (e.g., based on key performance criteria such as plant 2 000) after which any production receives only location and size. Differentiating payments by re- the market price. This has the effect of limiting source quality also supports economically viable windfall payments to developers in resource-rich development of diverse sites and avoids inflated areas, while still creating incentives to site proj- payments to developers in resource-rich areas. ects in the best locations. The risk of payments flowing to poor-perform- Design feature 2: ing projects can be reduced by attaching mini- Type of support for production mum performance requirements. Finally, paying out the incentive over time rather than in a single Once quantity is established, the next key deci- lump sum leaves room for program administra- sion is the type of production support. This can tors to adjust the level of payment in response to generally be classified based on how the pay- changing market conditions. ment is determined: whether it is: 1) not tied to the energy market; 2) tied to the energy market −− Production-based mechanisms or 3) tied to a separate market. Production-based remuneration schemes are Production incentives that are not based on a characterised by periodic payments based on a market are administratively defined, typically ei- generator’s actual production and, in some cas- ther through a calculation of the levelised cost of es, the market price. By tying payment to perfor- energy or by auction. Once the incentive level is mance these schemes create an incentive to max- established, RES generators enter into a contract imise energy production, meaning developers for a certain number of years and are paid on a naturally prefer the windiest/sunniest locations, per-kilowatt-hour basis for any production re- the most efficient components and the best sys- gardless of demand. FiTs are the prime example tem designs. of this type of flat, out-of-market approach. They The same cannot be said for the impact of pro- have been historically effective because they cre- duction support on operating behaviour. While ate an environment of high investor confidence the operating decisions of wind and solar are due to the elimination of market price risk. How- simplified compared to their thermal counter- ever, from a pure market design perspective, they parts due to their intermittency and zero vari- are the least compatible with markets: generators able cost, decisions such as shutting down in the are insulated from market signals and are there- presence of negative prices, or scheduling main- fore more likely to create distortions, such as con- tenance during off-peak times, are not driven by tinuing to produce even when prices are negative. price signals coming from wholesale energy mar- Support mechanisms tied to the energy market kets. Furthermore, linking support payments to require generators to sell their production on the production adds a second signal that may cause wholesale market and then provide a premium generators to change their behaviour, resulting in on top of the market price. These schemes are potentially distortive impacts across the broader broadly classified as feed-in premiums, though market. they have a variety of alternative names de- Production-based support mechanisms are char- pending on the design details. These schemes acterised by two key design features, as follows. are “market integrated” by nature, as generators are exposed to price signals and tune their op- Design feature 1: erating decisions accordingly. However, this does Amount of production receiving support not eliminate the potential for market distortions The first key decision in the design of produc- completely, as RES generators may still have an tion-based schemes is the quantity of genera- incentive to continue producing in the presence tion to support. Regulators can essentially decide of negative prices – at least until negative pric-

85 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

es exceed the size of the premium. The extent of average of market prices (e.g., hourly), then gen- these distortions critically depends on how the erators are continually “topped up” to the strike 2 premium is calculated and whether it is fixed or price and are thus well-insulated from market vol- “sliding”: atility. The support mechanism functions like a FIT and generators receive signals to produce until •• Fixed premiums are defined in advance and prices exceed the negative of the strike price; in therefore expose generators to the same lev- fact, regardless of when they produce and of the el of market volatility and price risk as any market price at that time, their revenue is always other market participant. The production tax reset to the strike price. At the other extreme, if credit in the United States, the “green bonus” the reference price is based on a long-term aver- in the Czech Republic and a similar scheme in age (e.g., yearly), then the premium is more sta- Slovenia are all examples of fixed premiums ble on an hour-by-hour basis, and the mechanism (Fraunhofer ISI, 2014). sends signals equivalent to a fixed premium, ex- •• Sliding premiums are calculated ex post as the posing generators to market volatility and price difference between a strike price (which reflects risk. In this case, the generator does have a pref- the long-term price needed to recover fixed erence for producing at high-price hours. In prac- costs) and a reference price (which reflects in tice, many countries choose a middle ground, bal- some way the payment received by participat- ancing revenue certainty with market exposure. ing through the electricity markets). By contin- Germany, for example, calculates reference prices ually correcting total payment to a fixed strike based on a monthly average while Finland uses a price, a sliding premium provides a high level of three-month average (RES-LEGAL, 2015). These revenue certainty while still exposing producers different approaches are represented graphically to market prices in real time. in Figure 2.28. The duration of the settlement period for apply- Finally, the sliding premium may be implementing the reference price is a critical factor conditioned as a purely financial contract, as in the case ing the behaviour of generators (Huntington et al., of the United Kingdom’s Contract for Differences. 2016). If the settlement is based on a short-term

Figure 2.28 Reference price settlement periods

Strike Price

Market Price

Premium based on long-term average Medium-term average Short-term average

Premium renumeration

Market renumeration

Eective Price Signal

Source: Adapted from Huntington et al., 2016

86 WHOLESALE MARKET DESIGN

Under this arrangement, generators are still Attempts to limit distortions through market de- “topped up” when reference prices are below sign “fixes” may create new problems. Banning strike prices (which is most of the time), but are negative prices outright removes the channel forced to pay back the difference when refer- through which base-load plants might contin- ence prices exceed strike prices (peak days). This ue operating, by effectively paying renewable works because generators still have an incen- generators to shut down. Another option is to tive to produce as prices are above their variable force RES generators to shut down when pric- cost, but their revenue is effectively capped at es are negative and simply pay them when they the strike price. are available. However, this creates a dangerous “cliff-edge” effect where generators are high- A third alternative for production-based schemes ly sensitive to small changes in market prices: is to tie premiums to an entirely separate mar- plants could suddenly switch between a desire to ket where the value of renewable generation is run at full output and access premium payments determined by the demand for it. These mecha- when prices are above zero and to turn off and be nisms are broadly categorised as “quota obliga- paid upon their availability when prices are below tions” linked with “tradeable green certificates”. zero. These oscillations could disrupt the broader While quota schemes expose producers to the energy market and complicate system operation efficiency of market prices, they provide less rev- (DECC, 2012). enue certainty for investors. Price floors are commonly introduced to address this (NREL, 2011), Production-based schemes will always suffer but by guaranteeing a minimum value for the from this fundamental tension between market premium quota, these schemes may cause some integration and limiting distortions because, one market distortions as traditional feed-in premi- way or another, they alter the price signals from ums. Even a well-functioning certificate market is the energy market. not immune to these problems. In a market with relatively stable prices (e.g., Sweden) generators −− Specific support mechanisms for prosumers may still find it in their interest to produce when Prosumers are electricity customers who both prices are zero or slightly negative, knowing that consume and produce electricity.36 The pros and they can still sell their production in the certifi- cons of support mechanisms for generators, as cate market for a profit. described in the previous section, are equally applicable to generation from prosumers. Sup- Assessment of market compatibility port mechanisms for prosumers have the add- Fixed premiums expose generators to efficient ed dimension of interaction with the retail tariff. market signals and introduce limited distortions, Depending on the metering technology and ar- but they come at the price of reduced revenue rangement, production can net consumption and certainty. Sliding premium schemes attempt to reduce consumption tariff payments. When vol- provide greater revenue certainty but create the umetric tariffs are bundled, prosumers have the potential for negative price distortions as genera- opportunity to reduce or avoid non-energy costs, tors bid up to the negative of the strike price. Cer- such as taxes, policy costs or network charges. tificate schemes are often perceived as the most This aspect has a more significant impact when “market compatible” since market signals and pre- prosumers are allowed to “bank” production mium determination are fully decoupled, but the credits and apply them as negative consumption risk introduced by the certificate market can be at a later date. significantly higher. Measures to reduce that risk, such as price floors, transform the scheme to one that functions like a traditional fixed premium.

36. This topic is addressed here from the perspective of these mechanisms’ compatibility with overall market functioning. In particular, we look at potential distortions introduced by the inappropriate design of support mechanisms. Chapter 3 will tackle these issues again from the perspective of their impact on the distribution network and the design of distribution and retail tariffs.

87 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

market can act as a basis for valuing grid injec- The metering technology plays an important role tions, thus providing a more efficient long-term in these support mechanisms. Meters must be bi- price signal (EURELECTRIC, 2015). 2 directional, therefore providing the possibility to distinguish between consumption and produc- Design feature 2: tion. Furthermore, smart meters with measure- Value of net excess generation ment intervals that are aligned with the intervals Net excess generation (NEG) refers to any proof the wholesale market allow the market value duction not consumed on-site and not converted of both injected and consumed energy to be de- to production credits to be treated as negative termined separately. This is especially relevant consumption. In net-metering schemes, NEG ex- when the two profiles are not aligned – for exists only when on-site production exceeds con- ample, with on-peak production (solar) and off- sumption over the course of the netting period. peak consumption (a night-oriented residential Schemes that do not allow banking of production profile). This also helps manage imbalances and credits effectively treat all grid injections as NEG. attributes imbalance costs more fairly (Eurelec- tric, 2015). The value of NEG influences both the operating and investment decisions of prosumers. If NEG is The remainder of this section focuses on two key valued at or above the retail rate, the incentives design features of support schemes for prosum- are equivalent to net metering: customers are in- ers: namely the parameters of the netting period, different to their rate of self-consumption. This and the value of excess generation. can also lead to the oversizing of generation as- Design feature 1: sets. On the other hand, if NEG is valued below Length and timing of netting period the retail rate, customers have a direct incentive to increase their rate of self-consumption and tai- The netting period refers to the window between lor the size of the generation systems accordingly. meter readings when customers can apply generation credits to their bill, effectively “netting out” Assessment of market compatibility their consumption. The longer the netting period, In any arrangement that allows for self-consump- the greater the potential subsidy since consumers tion, prosumers will have the opportunity to have more opportunity to use generation credits avoid paying the retail rate, and any associated to offset their consumption. Shorter netting peri- network costs. Many schemes allow prosumers to ods, combined with a lower rate for excess gener- carry their production credits forward and “net ation, create incentives for prosumers to increase out” their consumption at a later time. From a their rate of self-consumption. pure market perspective, schemes that do not al- Most states in the United States use an annual low the “banking” of production credits, or those netting period, though some, like Alaska, allow that minimise the netting period, lead to fewer credits to be carried forward indefinitely. Two distortions. Of course, this depends on whether utilities in California, PG&E and SDG&E, recent- the design of the network tariffs takes advantage ly proposed shortening the netting period from of the pattern of injections and withdrawals of yearly to monthly in an effort to control support the network users. costs and reduce cross-subsidisation of custom- Rather than allowing the banking of production ers with solar production (PG&E, n.d.). Most EU credits, support for prosumers may be delivered countries, including Belgium and the Kingdom of by adjusting the value of NEG. As long as the value Netherlands, , also use an annual netting period. remains below the retail rate, customers will have Denmark uses a very short hourly netting period incentives to make both efficient operating (in- (European Commission, 2015c). creased self-consumption) and investment deci- The timing of the netting period is another im- sions (proper sizing), leading to fewer market dis- portant dimension: by aligning the period with tortions and greater overall efficiency. But the only the trading interval of the wholesale market, the fool-proof approach is a complete redesign of the retail tariffs, as discussed in the next chapter.

88 WHOLESALE MARKET DESIGN

2.5. CONCLUSIONS Also, the balancing segment must be reformed AND RECOMMENDATIONS if RES resources are to become responsible for their imbalances. Balancing products should be In this chapter, we saw how wholesale market de- revised to unlock the potential of all flexible re- sign will have to evolve in the near future to ef- sources. Balancing capacity and balancing ener- ficiently integrate large shares of RES and DER. gy should be two separate products, procured The short-term market is probably the segment at different stages. Resources not cleared in as that requires the most adjustment. As many char- balancing capacity may be allowed to provide acteristics of the power sector become more vari- balancing energy if they can do it efficiently. Up- able, the time granularity of market signals has to wards and downwards reserves should be pro- be increased. Market time frames must be adapt- cured separately too since some resources may ed and become more flexible. be able to provide one product but not the other. Minimum-size requirements should be avoided to In terms of location, the deployment of variable the largest extent possible so as not to discour- technologies is likely to increase congestion on age RES and DER participation. the transmission network, which may exacerbate the potential disadvantages of zonal pricing when In the long term, we saw how capacity (or, more compared with nodal pricing. Nodal pricing pro- generally, generation adequacy) mechanisms are vides more accurate operation and investment climbing regulatory agendas in many liberalised signals, and its drawbacks in terms of long-term power systems. RES technologies can represent market liquidity and end-consumer discrimina- a valuable opportunity for system adequacy. De- tion may be reduced through financial transmis- pending on the characteristics of the power sys- sion rights and a proper tariff design. tem (especially on the share of hydropower in the generation mix), the variability of some RES A large deployment of renewable resources also technologies may not be an obstacle for these re- requires more complex bidding formats, if an ef- sources to provide reliability products efficiently. ficient economic dispatch is to be guaranteed. Bids may need to include an explicit representa- For this reason, RES resources should be allowed tion of most of the technical constraints of power to participate in generation adequacy, and also to resources. Also, bids may be tailored to certain be exposed to the corresponding market signals, products, such as those coming from demand re- which may result in the development of technol- sponse. ogies or strategies that maximise the availability of RES units during scarcity conditions. Harmoni- There are two key approaches to clearing bids and sation is called for in many areas, especially in the setting a price: that of the US ISO (dispatch-based calculation of a firm’s energy/capacity and in the pricing plus discriminatory pricing) and the EU PX application of penalties for underperformance. (price-based dispatch). The best option might be some point in the middle. An optimum clearing Finally, RES participation in capacity mechanisms methodology would find a balance between: 1) does not directly imply the elimination of any oth- uniform prices that may oversimplify the problem er form of RES support scheme. These two reg- and result in potential economic inefficiencies ulatory instruments can coexist – the only strong and 2) large uplifts that may distort the efficient requisite being that the incentive provided by the price signal. support scheme account for the remuneration earned on the capacity market. In the very-short-term, new solutions must be found to improve the reserve demand function, RES support mechanisms may be designed in increasing its elasticity to price. At the same time, several ways. The key question is whether to pro- the linkage between energy and reserve markets vide the incentive based on the capacity installed has to be enhanced in order for an energy market or the energy produced. Also in this regard, a bal- to properly reflect the value of reserves. ance must be found between optimal investment incentives and minimal market distortions.

89 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES 3

3.1 INTRODUCTION service cost may no longer be valid. The distribution system operator need to adopt advanced, in- Over the past decade, distribution networks have novative approaches to managing the distribution experienced the connection of an increasing grid. The active management of networks will help number of distributed energy resources (DER), solve network constraints and reduce relative- most of it distributed generation (DG), but also ly costly network reinforcements. This approach active demand response.1 In the near future, elec- requires distribution companies to interact more tric vehicles and distributed storage will have a closely with their network users and to deploy ad- substantial presence in several systems. The cur- vanced network technologies, which will play a rent – and anticipated – growth of DER demands crucial role in future power systems. The transition new approaches to operating and planning distri- towards smarter distribution networks should be bution networks, and the regulation of distribu- initiated through policies and regulations promot- tion and retail activities. ing innovation and demonstration projects. Sever- What are the major challenges anticipated? For al mechanisms have been used around the world the time being, most changes will be on the DG to help distribution companies efficiently integrate side. A high presence of DG may increase un- growing levels of DER through more active grid certainty around distribution networks’ remu- planning and operation. Examples of these mech- neration since the methods that regulators have anisms and of relevant demonstration projects are traditionally employed to estimate an efficient discussed in Section 3.2.

1. Distributed energy resources (DER) consist of small- to medium- scale resources that are connected mainly to the lower voltage levels (distribution grids) of the system or near the end users and are comprised of three main elements: distributed generation, energy storage and demand response. Demand response (DR), also known as demand-side management (DSM) or energy demand management (EDM) refers to the possibility to shift energy loads around in time. The management of small end-users must be achieved automatically at the user level, which requires online communications. In this regard, smart meters represent a key enabling technology of demand response. Distributed generation (DG) is a generating power plants serving a customer on-site, or providing support to a distribution network, and connected to the grid at distribution level voltages. For the purpose of this report, the technologies used for distributed generation are renewable energy technologies.

91 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

New regulatory approaches are needed for metering technologies are adopted. Section 3.4 the remuneration of distribution networks. The discusses the negative consequences of inap- 3 method currently used by most regulators relies propriate tariff structures in combination with on some multiplicative factor – validated by ex- self-consumption policies and discusses some perience – applied to the volume of distributed measures that regulators and policy makers have electricity. The network charges applied to the implemented to tackle such challenges. Several end-consumer are also, typically, volumetric, i.e., guiding principles for a truly cost-reflective tar- proportional to the amount of consumed ener- iff design and the sustainable development of gy. However, with high presence of DG, revenues self-generation on consumers’ premises are pro- may fall – both the total amount determined by vided in this section. regulators and that collected through the appli- Last but not least, owing to the presence of DER cation of network tariffs. This is in large part be- in their grids and the flexibility these services may cause self-generation implies a smaller amount provide, distribution companies will need to take of distributed electricity. In order to mitigate the on new roles as market intermediaries and sys- potential financial impact of this on distribution tem operators. This transformation means that, in utilities, some regulators have started to imple- the near future, distribution companies will need ment mechanisms that decouple the amount of to interact more closely with DER, which will be allowed revenues from the volume of delivered at least partly dispatchable. In this new context, energy. Also, some compensate companies for DER will provide services both to a distribution the incremental network costs. However, these company itself and to upstream stakeholders and short-term measures are not enough. More must markets. Since the distribution companies will be done to determine efficient cost levels and to need to act as mediators in these transactions, incentivise the degree of innovation required to e.g., performing technical validation or verifica- integrate high shares of DG efficiently. Profound tion of service delivery, the traditionally simple regulatory changes are required to encourage interactions between distribution companies and the deployment of more active grid planning and system operators – whether independent system operation. The regulatory changes to support DG operators (ISOs) or transmission system opera- integration and to promote long-term efficiency tors (TSOs) - ought to be revisited in order to en- are discussed in Section 3.3. sure the efficient and secure operation of a large- The development of generation on consum- ly decentralised power system. Thus, to some ers’ premises for self-consumption presents im- extent, distribution companies may become to portant benefits both for end users as well as retail markets and DER what transmission sys- for the distribution grid. However, as a result of tem operators are to wholesale markets and cen- tariff structures that do not adequately reflect tralised generation. Moreover, as part of their the true costs that each network user imposes newly acquired roles, distribution companies may on the system and outdated metering technol- be required to deploy new types of infrastructure ogies, self-consumption, especially when com- – such as advanced metering infrastructure (AMI), bined with net-metering policies, may jeopardise energy storage or electric vehicle (EV) charging the financial viability of the system. Acknowledg- stations – that do not fall within their convention- ing this, regulators have already started to apply al core activities. Section 3.5 addresses the new mechanisms to mitigate the problem in the short roles to be adopted by distribution utilities as sys- term. However, the fundamental issues will not be tem operators and market facilitators, including solved until cost-reflective tariffs and advanced in the deployment of infrastructure.

92 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

3.2 ADVANCED SOLUTIONS FOR work costs (investment, maintenance and ener- DISTRIBUTION NETWORK gy losses cost) compared to a business-as-usual PLANNING AND OPERATION approach (see, for example, Cossent et al., 2011). Box 3.1 summarises the results of analyses aiming at quantifying the effect of DG deployment A growing penetration of distributed energy on network costs, the level of investments to ac- resources, particularly distributed generation, commodate DG integration as well as the savings requires innovative approaches to distribution that might be achieved through innovative tech- network planning and operation. nologies (such as storage) and more active network operation. The presence of generation in distribution net- Policy makers all over the world acknowledge works is beneficial, thanks to the fact that DG is that not co-ordinating the development of DER located near end-consumers. It can partially off- and distribution grids is inefficient: set demand growth, reduce peak load and allow distribution companies to defer grid reinforce- “Thus far, there has been limited incorporation ments or reduce energy losses. However, in many of demand response and energy efficiency into cases the absence of locational network charges, distribution system planning efforts, and very as well as policy design, do not encourage effi- little incorporation of distributed generation. cient operating decisions regarding generation […] System planners are appropriately conser- location. And even if the correct locational sig- vative, and inclined to consider only resources nals were sent, operators can make operating that are well known and can be relied upon […] decisions based on many other factors. Thus, in These challenges, however, should not preclude practice, DG is not necessarily located close to consideration of available, feasible, and cost-ef- the loads or operated efficiently. Moreover, its fective DER solutions as part of any distribution production is largely variable and may not coin- system planning efforts” (New York DPS, 2014: cide with times when local demand is highest.2 14–15). Thus DG production does not necessarily take “[R]egulatory frameworks shall enable distribu- place where and when it is most needed from a tion system operators to procure services from network perspective. resources such as distributed generation, de- Another issue to consider is that distribution mand response or storage and consider ener- companies rarely rely on DER to optimise net- gy efficiency measures, which may supplant the work planning – and reap its potential benefits. need to upgrade or replace electricity capacity Instead, they usually follow a so-called fit-and- and which support the efficient and secure op- forget approach to DG connection, which con- eration of the distribution system” – Proposal sists of reinforcing the grid at the time of con- for a revised Electricity Directive, Art. 32.1 (Eu- nection in such a way that any future operational ropean Commission, 2016b). problem is prevented. This means that network To achieve these goals, distribution utilities should capacity by itself must be sufficient to cope not adopt a more active role in distribution network only with peak demand conditions, but also with planning and operation. Implementing advanced peak net generation, even if these conditions only solutions often requires utilities to take the risks last for a few hours per year (Eurelectric, 2013). inherent in using new technologies. Pilot projects Consequently, distribution costs can actually in- ought to be encouraged through funding mech- crease significantly due to the connection of DG. anisms that partly hedge distribution companies Several studies have demonstrated that the im- against these risks. This would help ensure that plementation of active network management suitable solutions are available when needed. helps lower the impact of DG on distribution net-

2. The coincidence in time of local generation and demand is a key factor determining the impact of DG/RES on the distribution network. For instance, solar PV production may yield significant benefits in areas dominated by commercial consumers with a significant air-conditioning demand. However, it is not possible to draw general conclusions about DG impacts since these depend on many factors, such as location, season, the year load composition, etc.

93 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 3.1 Network investments and advanced solutions to integrate distribution generation 3 Figure 3.1 presents the outcome of a study assessing the In 2011, the UK Department of Energy and Climate Change influence of photovoltaics (PV) on distribution network and the energy regulator – the Office of Gas and Electric- cost and energy losses for 12 prototype networks. These ity Markets (OFGEM) – created the so-called Smart Grids networks were constructed using a distribution network Forum for public authorities, industry representatives and planning model, called reference network model, and stakeholders to exchange ideas about the future of their corresponding to different reference locations across the energy networks. Among other initiatives, a forum study United States, selected based on solar radiation levels and evaluated the future investment that distribution net- population density. The vertical axis shows the increase works need to accommodate expected DER by the year in network costs driven by solar PV, relative to a scenario 2050 under different scenariosa (EA Technology, 2012). without DG. The horizontal axis measures PV penetration Figure 3.2 shows how, under all the scenarios, the consid- levels, defined as the share of the annual electricity de- eration of smart grid investments (either in a top-downb mand supplied by local solar generation. In addition, the or incrementalc approach) yielded much lower costs than contribution of energy storage to the mitigation of PV in- a business as usual (BAU) strategy. The vertical axis rep- tegration costs is evaluated for different values of a stor- resents the net present value of expected network costs age factor. This parameter represents the reduction in DG throughout the period 2012–50 quantified by means of a peak injection achieved by the storage system. A value distribution-planning model for a set of scenarios. Each of zero means that there is no reduction in the maximum scenario displayed in Figure 3.2 corresponds to a differ- annual power injection, whereas a value of one indicates ent penetration level of low-carbon technologies (solar that storage fully offsets PV injections. The results show PV, heat pumps and electric vehicles). that when storage reduces the peak PV production seen by the grid, significant amounts of PV can be connected Figure 3.2 Expected network investments in Great Brit- without a significant network cost increase. ain to accommodate low-carbon technologies connected to the distribution network under different planning scenarios (2012-50)

Figure 3.1 Effect of energy storage to mitigate Billions the impact of PV penetration on distribution costs 30

25 2.0

1.8 20 1.6 15 1.4

1.2 10

Network and Losses Cost and Losses Network 1.0

Relative Change in Total Yearly Change in Total Relative 5 0 5 10 15 20 25 30 35 40 (2012-2050) cost network of distribution NPV PV Energy Share (%) 0 Scenario 0 - Scenario 1- Scenario 2 - Scenario 3 - Hight Medium Medium Low SF=0.0 SF=0.2 SF=1.0 (Low DSR) BAU Incremental Top-Down Source: MIT, 2015 SF stands for storage factor Source: EA Technology, 2012

(a) Scenario 0 - high domestic decarbonisation; Scenario 1 - domestic decarbonisation to meet carbon targets; scenario 2 - domestic decarbonisation to meet carbon targets with low DSR; and scenario 3 - low domestic decarbonisation. (b) Top down approach - A strategy where an upfront investment of enabler technologies is deployed in advance of need, followed by investment as and when networks reach their headroom limits. (c) Incremental approach - A strategy where investment only occurs as and when networks reach their headroom limits. Enabling technologies are deployed alongside the solution variants on an incremental basis.

94 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

3.2.1 Proactive grid connection quent requests would trigger reinforcements. In approach and long-term case generators are obliged to bear these costs, network planning this could make later projects unviable. Moreover, an incremental reinforcement strategy can lead to inefficient network development due to the The management of grid connection existence of economies of scale in network com- applications should be reviewed to speed-up ponents. DG connections and allocate grid capacity more efficiently. Therefore, to the extent possible, a co-ordinated Information disclosure about grid hosting aggregate management of connection requests capacity and connection process should should replace the first-come, first-served ap- be promoted. proach conventionally applied. Thus, the allocation of existing network capacity and cost allocation would be carried out in a much more efficient Mitigating the effect that high levels of DG can and transparent way. See, for instance, the case have on costs of traditional distribution networks, of Ireland, which introduced a so-called “group and making the most of the benefits provided by processing”, described in Box 3.2, which has al- DER, require new approaches to grid connection lowed the handling of a large number of wind and network planning. Additionally, as discussed power connection requests.3 in Section 3.2.2 more flexible grid connection and planning methodologies require distribution Facing a similar problem related to the large companies to adopt a more active grid operation. number of network connection applications from promoters of generation projects, the Italian reg- DG units have conventionally been connected to ulator implemented a specific regulatory mecha- distribution networks using a fit-and-forget ap- nism to tackle the so-called virtual grid conges- proach. This means that, due to distribution com- tions, also described in Box 3.2. This approach is panies’ lack of control over DG units and DER in based on setting additional economic or admin- general, any potential operational problem, even istrative requirements before granting a certain exceptional events, had to be prevented through amount of grid capacity to applicants. The main network reinforcements at the time of connec- goals of the regulator were to prevent opportu- tion. This conservative strategy aims to ensure nistic behaviour from promoters and handle the that no operational problem will arise. While this very large volumes of grid connection requests. goal is a worthy one with low levels of DG, it may generate an excessive cost burden in the long run New generators wishing to connect to the distri- when high shares of DG are installed. This ap- bution grid typically have to submit an applica- proach ought to be progressively abandoned as tion to the distribution company specifying one DG penetration levels grow to prevent undue in- or more potential points of connection. The gen- creases in distribution costs and long lead times erator, meanwhile, normally does not possess any in DG connections. prior knowledge of network conditions. As a result, applicants may find their connection being First, grid connection processes should be re- delayed or subject to high connection fees due to visited. Network connection requests have con- insufficient network hosting capacity. ventionally been managed through a first-come, first-served approach, on a case-by-case basis. As a result, the first applicants would make use of the existing connection capacity, while subse-

3. This approach successfully addresses the problem of handling a large number of connection requests arriving at different times. As is often the case with regulations, no solution is perfect. The drawback of this approach is that it may impede the chances of better performing generators to compete with the existing ones, when their application is made later. It may also encourage the hoarding of connection permits, unless strict measures are taken to prevent this behavior.

95 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 3.2 Regulatory mechanisms to handle a large number of DG connection applications 3 Group processing of grid connection applications projects amounting to 128 gigawatts (GW) and 22 GW in Ireland had applied for connection to the transmission and distribution grid, respectively, whereas the total installed At the end of 2003, Ireland’s energy regulator instruct- capacity and peak demand were roughly 111 GW and 57 ed network operators, both transmission and distribution GW, respectivelya. Furthermore, generation projects were companies, not to accept any new grid connections due concentrated in areas where the grid presented less host- to concerns about the secure operation of the system. ing capacity. Hence, important virtual grid congestions Nonetheless, the number of connection requests from were identified, i.e., potential grid constraints that would wind power plants kept increasing during this moratori- occur in case the existing DG connection requests would um. Thus, the regulator decided to implement a “group materialise. Two main mechanisms were devised to ad- processing” with the following goals: dress this challenge for non-residential consumers/prosumers in critical network areas, i.e., those facing prob- “Securing early system access has become lems of virtual congestion: particularly valuable and important to applicants. This also mean that putting in place an orderly regime −− Provide the right for network connection and access for the resumption of connection offers is essential after additional economic guarantees had been given to realizing Ireland’s renewable energy obligations.” to the corresponding grid operator. The amount of such (CER, 2005) guarantees would be proportional to the network capacity requested. The guarantees would only be returned This method consists of jointly processing connection if the projects were realised and connected within two applications from renewable generators larger than 500 years from the time of application. kilowatts (kW) in subsequent batches or “gates”. In each Provide the right for network connection and access gate, existing applications are sorted into groups. Gen- −− only to those projects that had already obtained con- erators in the same group are those sharing the use of struction and operation permits. Under this alternative, a portion of the grid. Then each group is evaluated by promoters have specific deadlines to begin the construc- its corresponding network operator, both technically and tion work of the plant before losing the right to connect, economically. In case this results in a positive evaluation, which depends on the voltage level. applicants receive a connection offer, including their corresponding share of grid reinforcement costs, which are The application of the first solution, initially preferred, allocated in proportion to the capacity of each plant. was hampered by a litigation process. Hence, the second alternative was finally adopted. The implementation pro- Mitigating virtual congestion in Italy cess is described in AEEGSI (2011; 2012a; 2012b). Residen- The Italian regulator published a consultation document tial consumers/prosumers were exempt from both these concerning new technical and economic grid connection requirements. requirements for generators in response to the large amount of new generation capacity that had requested a grid connection (AEEGSI, 2010). By the end of 2010, a. http://www.autorita.energia.it/it/docs/11/073-11arg.htm

In order to achieve faster and more transparent tion charges. This would, in turn, encourage gen- grid connections, distribution companies should erators to request their connection at the point be obligated to disclose information on avail- where network conditions are most favourable. able. On the publication of available generation Box 3.3 provides several examples of distribu- hosting capacity at potential points of connection utilities that publish this type of information tion would allow DG promoters to estimate in ad- in different forms, sometimes subject to registra- vance whether their application will be successful tion. The formats range from simple online calcu- or what location would result in lower connec- lators to interactive maps using geographic information system (GIS) technology.

96 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Box 3.3 Network information disclosure for new DG connections

In 2012, The United Kingdom’s energy regulator, the Of- Likewise, distribution utilities in California were mandated fice of Gas and Electricity Markets (OFGEM) decided to to elaborate a so-called “distribution resources plan”. This host several discussion forums where distribution com- sought to identify optimal locations for the deployment of panies and generators could discuss the increasing num- distributed energy resources (DER). The three main dis- ber of generators seeking to connect to the distribution tribution utilities in the state have published the detailed grid, the difficulties they faced when going through the information used for such analysis.a Figure 3.4 shows an connection process and the possible solutions. Today, interactive map where users may check the available ca- distribution companies in the United Kingdom provide pacity for generation connection in different network ele- extensive and easy-to-interpret information about the ments. As in the previous case, colour-coding and pop-up connection process, as well as detailed information about menus are used to display the relevant information. the conditions of the grid. Available tools include distributed generation (DG) heat maps (SP Energy Networks), Figure 3.4 DER integration map in California generation availability maps (Northern Power Grid) and a (Southern California Edison) DG mapping tool (UK Power Networks) (see Figure 3.3).

These are interactive web-based tools that allow potential applicants to perform a preliminary assessment of grid-hosting capacity in different locations throughout the distribution grid. Figures below show two examples of map-based information provided by two distribution companies in the United Kingdom. A colour code is used to indicate the loading level of different network elements, i.e., feeder circuits or substations. More detailed information (voltage level, rated capacity, existing DG capacity connected, remaining hosting capacity, etc.) is displayed by means of pop-up menus. Source: Southern California Edison, 2017 a. See http://www.cpuc.ca.gov/PUC/energy/drp/index.htm.

Figure 3.3 Examples of map-based information for potential DG connections in the United Kingdom

This map shows 33/11KV Primary · Transmission Constraint Substations and a geographical · General Capacity Connected footprint of the areas they feed. · Reverse Power Flow they are coloured red, amber or · 33kV Fault Level green depending on whether any of · Percentage of 11kV Circuits the following have been identified Saturated as affecting further generation connections.

SUBSTATION INFORMATION

Name: Fourstones Primary Classification: AMBER Nominal Voltage (KV): 20 (Latitude, Longitude): (55.0119.0-‐2.165)

Fault Level 52.69% (%): Voltage constraint issue: 5 Reverse Power Flow Capability: 100|40 (MVA) Upstream: AMBER LIMITED SOURCE SUBSTAION CAPACITY Physical Constraints: GREEN

VIEW DETAILS

DG heat map Generation availability map Source: Adapted from SP Energy Networks, 2017 Source: Northern Power Grid, 2017

97 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Over the long term, a more comprehensive, 3.2.2 From reinforcement-based co-ordinated and forward-looking network plan solutions to an active network 3 ought to be undertaken (Jamasb and Marantes, operation 2011; Jendernalik, 2015). Conventionally, this exercise broadly consists of forecasting peak de- Integrating DG efficiently requires the mand over the planning horizon to determine adoption of active network management as the necessary grid reinforcements. However, the an alternative to simple conventional grid times required to install a DG plant are usually reinforcements. shorter than those required to reinforce the grid. Distribution companies will increasingly need to foresee the connection of DG well in advance in Capturing the potential benefits of demand re- order to prevent potential bottlenecks (Jamasb sponse and DG in distribution network planning and Marantes, 2011). necessarily involves implementing a more active network management. Thus, instead of prevent- The main added difficulty is that planning net- ing grid constraints by simply reinforcing the networks with large shares of DG, typically variable, work or delaying the connection of DG, distribu- requires knowing not only how much capacity will tion companies may solve congestion or voltage be connected, but also when and where DG will problems during day-to-day operations. be using this capacity. To do this, utilities will have to move towards a probabilistic scenario-based On the one hand, distribution companies will grid planning process. More detailed geographi- need to adopt an enhanced use of information cal information, such as renewable resource avail- and communication technologies (ICT) and inability or land-use mapping, should be employed novative systems to solve network constraints. to quantify existing DG potential. Lastly, both These solutions comprise, among other systems, network reinforcements or expansions and ad- automated voltage control to mitigate the volt- vanced grid technologies should be considered age rise effect caused by DG (see Box 3.4) or au- by the distribution companies in order to deter- tomatic grid reconfiguration to reduce the load- mine efficient and robust investment strategies. ing of a distribution feeder by transferring part of the DG feed-in to a neighbouring one. Regulation can promote this change in planning practices by mandating that distribution compa- In addition to deploying innovative technolo- nies submit justified investment plans following gies, utilities should interact more often with DER common criteria. An example of such practices to efficiently manage network constraints. The is that of the United Kingdom, where distribu- most straightforward approach is mandating DG tion companies have to apply a common meth- units to comply with certain communications re- odology to justify their business plans based on quirements and dispatch signals sent by the dis- a benefit-cost analysis, including innovative grid tribution system operator (DSO), as in the exist- solutions. In this regard, the regulator states ing PV curtailment system in Germany. A similar that distribution companies shall use a techno- approach may be found in the United Kingdom, economic model called the Transform Model4 where distribution companies may temporarily (or other modelling tools with comparable capa- curtail those consumers connected at the high bilities) and follow the long-term scenarios de- voltage level who have agreed to be curtailed fined by the Department of Energy and Climate when needed in exchange for lower network Change (OFGEM, 2013a). charges (Box 3.5).

4. The Transform Model is a spreadsheet model which has been developed and is property of the consultancy firm EA Technologies. Further information can be found at http://www.eatechnology.com/products-and-services/create-smarter-grids/transform- model%C2%AE.

98 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Box 3.4 The need for more active network management: voltage control with DG

In the absence of distributed generation (DG), bus volt- The conventional solution to this kind of problem con- ages drop along the distribution feeders. Provided that sists of permanently lowering the voltage at the substa- the network is adequately designed, no voltage problems tion and modifying the transformer tap position until the should occur. Nonetheless, the increasing connection of voltage is kept within margins. By doing this, as shown DG to this network could raise bus voltages up to the in the right-hand side of the figure, all bus voltages are point where the allowable upper bound is violated. For shifted downwards. However, because voltages are not instance, the left side of Figure 3.5 shows the voltage pro- normally monitored in real-time, and tap positions cannot file throughout a distribution feeder under different load- be adjusted accordingly, this could lead to an excessive ing scenarios with and without DG. It can be seen that the voltage drop in a scenario with high load and very low DG presence of DG, particularly under low-load conditions, feed-in (right part of figure). Hence, as DG penetration can cause an increase in voltages that at times may ex- grows, the distribution company would increasingly need ceed allowable voltage limits, represented by a dashed to monitor local voltage levels and dynamically adjust red line, in some buses (usually those closer to the loca- the substation voltage depending on actual, rather than tion of DG units). expected, network conditions. Therefore, voltage control approaches should evolve from static passive solutions to location and time-dependent control strategies based on local measurements and grid monitoring.

Figure 3.5 Impact of DG on distribution voltage control strategies

Voltage Voltage Low load Upper Upper Voltage Violation Permitted High DG feed-in Permitted Low load voltage voltage High DG feed-in High load High load Low DG feed-in Low DG feed-in Set value Busbar voltage Set value primary Busbar voltage substation primary substation Low load No DG Low load Lower Lower No DG permitted High load permitted No DG High load voltage voltage Voltage Violation No DG

Network length Network length

Source: Adapted from Jendernalik, 2015

Mechanisms that allow the distributor to tempo- Existing connection requirements and planning rarily curtail the power injection or withdrawal of criteria may not be flexible enough. Thus, rules an end user for security reasons are referred to obliging distribution companies to size their grids as variable or non-firm network access (Eurelec- according to a worst-case scenario should be tric, 2013) or smart contracts (EDSO, 2015). There modified to allow distribution companies free- is a trade-off between: 1) reducing the value of dom to decide whether to reinforce the grid or the RES due to curtailment and 2) the benefits offer non-firm access contracts to their users. reaped in terms of swifter DG integration. Thus, A limitation of these connection agreements is solutions should be promoted, with the aim of that the compensation for the provision of flex- keeping the amount of curtailed DG production ibility is related to the level of network charges, low, preventing costly grid reinforcements, in- which does not necessarily reflect the actual value creasing network hosting capacity, and allowing that the service has for the distribution company. for a faster DG connection and access. Moreover, they do not generally allow different DER to compete for the provision of the service.

99 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 3.5 Non-firm distribution network access as a means to prevent network constraints 3 PV curtailment by distribution system problems and reducing lead connection times, the mech- operators in Germany anism does not allow distribution system operators to entirely rely on distributed generation (DG) management as According to the German Renewable Energy Sources Act, an alternative to network reinforcements. or EEG, since January 2012, German distributors are entitled to limit remotely the injection of photovoltaic (PV) Demand-side management agreements installations above 30 kilowatts (kW), subject to compen- in the United Kingdom sation. Plants below 30 kW may choose to permanently Distribution companies in the United Kingdom may of- limit their power injection to 70% of the nameplate ca- fer non-firm network access to their network users. For pacity (leaving the remainder to be curtailed or locally instance, Scottish Power offers their extra-high-voltage stored/consumed) or to install the same communication consumers a demand-side-management agreement. This system installed on larger plants. contract offers eligible consumers reduced charges in This mechanism should be applied only on a temporary exchange for their agreement to reduce their maximum basis in case of an emergency (congestion or overvolt- consumption in those periods, determined by the distri- age). The distribution company is still expected to eventu- bution company. Moreover, the demand charge will be ally reinforce the grid so that no curtailment is necessary. reduced in proportion to the amount of capacity that is Thus, while helpful in addressing short-term operational subject to such a reduction (SP Energy Networks, 2014).

Thus, more flexible and ad hoc approaches to en- gration is not the only factor driving smart grid able the interaction of DSOs with DER have been adoption. Additional motives include improved proposed (see, for example, CEER, 2015a; CEER, market functioning, the development of electric 2015b; and Eurelectric, 2013). These mechanisms, mobility, a reduction in energy losses and en- as discussed in Section 3.5, may take the form hanced network resiliency and reliability. of “bilateral flexibility contracts” (CEER, 2015b) Without specific measures in place, distribution or that of local markets operated by the distri- companies have little or no incentive to innovate bution utility and participated by local DER (Tre- since innovation involves increasing their risk ex- bolle et al., 2010; Eurelectric, 2013; Poudineh and posure. Regulations may not allow them to use Jamasb, 2014). In both cases, the service provider tariffs to recover the costs of innovative technol- may be the network user or an intermediary, such ogies; and the solutions adopted may not yield as a supplier or an aggregator. the expected results, leading to stranded costs. 3.2.3 Paving the way through Thus, it is up to policy makers and regulators to demonstration and pilot projects kick-start the transformation. Pilot projects should be promoted by means of Policy and regulation should promote ad hoc policy and regulation so that distribution the implementation of pilot projects and the companies are able to test innovative solutions exchange of lessons-learnt and best practices. while mitigating their risks in case of failure. Rele- vant regulatory instruments offer a higher rate of New technology deployment should be return on innovative investments or accelerated supported through the creation of depreciation methods, or allow the pass-through public-private collaborative networks. of demonstration costs or foster direct project funding through competitive schemes (CEER, More active distribution grids, largely reliant on 2015b). In recent years, numerous demonstration ICTs, are necessary to ensure an efficient integra- and pilot projects have been launched to pro- tion of DG. The term “smart grid” is widely used mote smart distribution grids worldwide. to refer to this new paradigm. Renewable inte-

100 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Joint public-private partnerships and fora pro- projects (both research and development, and mote the exchange of lessons learnt and identi- demonstration projects), amounting to Euro 3.15 fy common priorities. Examples of such collabo- billion in investments within the period 2002-14 ration networks at the regional level include the (JRC, 2014). The distribution of the budget devot- European Union (EU) Smart Grid Task Force,5 the ed to demonstration projects shown in Figure 3.6 US GridWise Alliance6 and the Indian Smart Grid clearly illustrates the priorities of enhancing net- Forum.7 Through these networks, the European work management, empowering consumers and Union, the United States, and India illustrate dif- integrating DER. ferent approaches and priorities in the transition Furthermore, a shift in the allocation of resources towards smarter distribution grids. At the glob- from research and development to demonstra- al level, the International Smart Grid Action Net- tion projects can be observed over the last few work (ISGAN),8 includes stakeholders from 25 years (Figure 3.7). This denotes the need to im- countries across five continents. plement demonstration projects to prove smart In the European context, smart distribution grids grid technologies with a view to large-scale de- are considered essential to achieve more-efficient ployment. grid operation, the integration of large shares of The American Recovery and Reinvestment Act of renewable generation and a fully functional retail 2009 (US Congress, 2009), in short the Recov- electricity market. The European Commission, ery Act or ARRA, represented a strong stimulus through its Joint Research Centre, has built up a for the development of smart grids in the United database comprising more than 450 smart grid States. This law allocated USD 4.5 billion to the

Figure 3.6 Distribution of budget per country and smart grid application in Europe, considering only demonstration and deployment projects

500 Smart Metering 450 Integration of large scale RES 400 Other Aggregation (Demand Response, VPP) 350 Integration of DER Electric Vehicles and Vehicle2Grid applications 300 Smart Customer and Smart Home 250 Smart Network Management

200 Budget [million Euro] Budget 150

100

0 UK FR DE ES IT NL DK BE SE AT PT GR CZ SI IE FI NO CH HU LT PL HR SK RO BG LV CY LU EE MT

Source: Adapted from JRC, 2014

5. http://ec.europa.eu/energy/en/topics/markets-and-consumers/smart-grids-and-meters/smart-grids-task-force. 6. http://www.gridwise.org/index.asp. 7. http://indiasmartgrid.org/. 8. http://iea-isgan.org/.

101 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

3 Figure 3.7 Cumulative annual budget of smart grid projects in Europe, 2002-14 3,000

2,500

2,000

1,500

1,000 Budget [million Euro] Budget

500

0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Total budget R&D budget Demonstration & Deployment budget

Source: Adapted from JRC, 2017

United States Department of Energy (US DOE) to In several Indian states, smart distribution grids promote the modernisation of the US electricity offer a way to tackle pre-existing challenges, system. Two major programs to foster demon- particularly high losses and poor reliability, and stration activities in the smart grid area were cre- to facilitate distributed generation, especially the ated: the Smart Grid Investment Grant and the rooftop solar generation, by allowing movement Smart Grid Demonstration Programs. These two and measurement of energy in both directions, programmes accounted for more than 93% of the using control systems and net metering that will ARRA funds for electric grid modernisation; over help “prosumers” (NSGM, 2017). In recognition of USD 3.4 billion and USD 0.68 billion, respectively. this, the Indian Ministry of Power has selected 14 They partially funded more than 130 demonstra- pilot projects led by different public distribution tion projects overall.9 Figure 3.8 shows the num- utilities (see Table 3.1). These demonstration proj- ber of demonstration projects addressing each ects mainly focus on improvements in metering type of functionality using data extracted from a systems (essential for any loss-reduction strate- database of smart grid demonstration projects in gy), peak-load reduction and faster recovery of the United States. supply after an interruption. In 2015, the government also launched the National Smart Grid Mis- Many of these projects tested more than one sion (NSGM), under which four additional proj- functionality. For instance, projects dealing with ects have been launched in 2016 (NSGM, 2017). advanced metering infrastructure (AMI) system Smart grid infrastructure investment is estimated frequently address pricing and consumer sys- to be about USD 45 billion over the period 2017- tems as well. This figure illustrates that the prior- 27 (T&D World, 2017). ities were mainly placed on improving consumer information, demand response and network effi- As discussed, in order to integrate large DG levels, ciency and reliability. As compared to the EU ac- distribution companies will need to implement tivities, the US projects focus less on retail mar- more advanced grid technologies and operation- kets than on grid resiliency and security. al solutions. This entails a deep transformation in current practices, which should be accompanied by suitable regulations encouraging the change.

9. Full information on the projects can be found at: https://www.smartgrid.gov/ and http://energy.gov/oe/information-center/recovery-act.

102 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Figure 3.8 Categorisation of US smart grid demonstration projects per type

Customer Systems 79

Advanced Metering Infrastructure 77

Pricing 76

Distribution Systems 69

DER 43

Transmission System 22

Energy Storage 16

Phasor Measurement Unit 14

Consumer Behaviour 9

Equipment 2 0 10 20 30 40 50 60 70 80 Number of projects

Source: Based on Smartgrid, 2017

Table 3.1 Overview of smart grid pilots in India

Functionalities Consumers Jurisdiction involved AMI AMI PLM OMS PQM DG Industrial Residential

Telangana 11,904 X X X X X

Assam 15,000 X X X X X X

Karnataka 24,532 X X X X X X

Chhattisgarh 1,900 X X

Puducherry 87,031 X X

Himachal 650 X X X X Pradesh

Jaipur 34,752 X X X

Kerala 25,078 X

Maharashtra 29,997 X X X

Punjab 9,818 X X

Tripura 46,071 X X X

Gujarat 39,422 X X X

Haryana 11,000 X X X

West Bengal 4,404 X X X

Notes: AMI – advanced metering infrastructure, PLM – peak load management, OMS – outage management system, PQM – power quality management, DG – distributed generation. Source: Based on NSGM, 2017 103 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

3.2.4 Conclusions as some broad conditions to prevent discrimina- and recommendations tory behaviour from the utilities. 3 Management of grid connection applications Despite the need for smarter distribution grids, should be reviewed to enable a timely and at present, distribution companies have scarce cost-effective DG connection. First, connection incentives to change, due to the inherent tech- requests should not be processed as they arrive, nology risks and the absence of suitable regula- but in a more co-ordinated way, e.g., in batches tory mechanisms. In the early stages, policy and per network area. This would result in a less dis- regulation should promote the implementation criminatory network capacity allocation and less of demonstration projects and the exchange of costly network reinforcements. Second, the pub- lessons learnt and best practices among the rel- lication of detailed information on the capability evant stakeholders. Thus, both the introduction of the existing network to host additional gener- of explicit economic incentives for pilot projects ation capacity enhances transparency and allows and the creation of public-private collaborative promoters to make better informed connection networks are recommended. applications. Overall, this should result in less need for reinforcements and swifter connection 3.3 ADVANCED REGULATION processes. TO ENCOURAGE INNOVATION Planning the distribution grid under high DG AND EFFICIENT DISTRIBUTED penetration levels becomes a more complex GENERATION INTEGRATION exercise. Greater uncertainty exists about DG location and future feed-in. Moreover, distribution companies will need to consider how advanced Conventional remuneration formulas and cost grid technologies contribute to the reduction assessment methodologies are ill adapted of conventional investments on the basis of to the high presence of DER. Efficiently cost-benefit analyses. Regulation can push this integrating high shares of DER requires transformation by mandating that utilities submit a major regulatory overhaul. detailed business plans compliant with some common format or methodology as part of the Distribution regulation has traditionally focused revenue regulation process. on ensuring adequate investment levels and pro- Integrating DG efficiently requires relying on a moting short-term efficiency gains whilst pre- more active network operation to solve network venting a deterioration in service quality. These constraints close to real-time, instead of exclu- priorities, when supported appropriately, proved sively preventing them through network rein- successful in a context of stable technologies and forcements. This involves the deployment of ad- predictable demand. vanced grid technologies and ICTs as well as a Under this regulatory paradigm, distribution closer interaction between the distribution com- companies require significant investments to ac- pany and the existing DER. Nowadays, only a few commodate large levels of DG under safe condi- countries allow this, mainly by means of flexible tions. Moreover, since distribution remuneration connection contracts which enable distribution is often linked to the amount of energy distribut- companies to temporarily curtail the power of ed, high levels of local self-consumption may re- a network user in case of an emergency. None- sult in a decrease in distribution companies’ rev- theless, as the presence of DER grows, more ad- enues. Consequently, utilities may see DG as a vanced forms of contracting, such as bilateral threat to their activities, and this could hamper its agreements or market-based approaches, should timely connection. Therefore, in the absence of an be adopted. This should be regulated by means of in-depth regulatory reform, regulatory measures newly developed distribution network codes de- should aim to at least neutralise the negative ef- fining the service to be provided by DER as well fects of DG on distribution revenues.

104 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

In addition, as discussed in the previous section, framework should encourage distribution com- a business-as-usual approach to distribution grid panies to carry out the most cost-efficient invest- planning and operation is not suitable to accom- ments, including smart grids. In the process, it modate large levels of renewable and distributed would ensure that remuneration is aligned with generation in a cost-efficient and timely manner. the companies’ financing needs and reward them Therefore, distribution companies should be en- partly on the basis of their performance, rather couraged to deploy innovative grid technologies than only on their efficiently incurred costs. and benefit from the flexibility offered by DER Another fundamental aspect that regulators need when this results in lower costs than conventional to take into consideration is that the changes that copper and iron reinforcements. In the early stag- advanced power systems are experiencing, from es, pilot demonstration projects are necessary to the increasing VRE penetration to a more active test innovative solutions and identify the most participation of consumers, are strictly interlinked promising ones on the basis of benefit-cost as- with power system digitalization. The evolving role sessments. of information and communication technologies The long-term transformation of the distribu- for the efficient management of the power system tion network will be achieved only with a major calls therefore for a close cooperation between regulatory overhaul. An appropriate regulatory energy and telecom regulators (see Box 3.6).

Box 3.6 Digitilisation and cooperation between energy and telecom regulators

Digitalisation can be defined as the transformation of launched a survey on “machine-to-machine” (M2M) tele- power system onto a two-layer system that combines the communication services. The energy regulator (AEEGSI) traditional electrotechnical technology with a new lay- contributed to the survey by collecting the experiences er of information and communication technology (ICT), and lessons learnt in the energy sector for the deploy- which provides functionality for remote monitoring, con- ment of smart grids and smart meters, both in the elec- trol and protection. Of course, this kind of schematisation tricity and gas sectors. The following aspects were high- does not mean that power system has not yet been dig- lighted in that context (AEEGSI, 2014). italized at all. System defense tools, like automatic load 1) Ensuring the necessary level of interoperability (be- shedding in case of severe frequency perturbations, were tween devices built by different companies, between introduced decades ago; and in the first years of the sec- systems based on different technologies, between data ond millennium, digitalisation made a further step ahead collection and management platforms, etc.) and easy in several systems, introducing, for instance, the remote switchability among different providers (for instance control of distribution networks. Nowadays, digitaliza- through e-SIM, which can be provisioned “over-the-air”); tion of power systems has much to do with deployment of smart grids and intelligent metering systems and with 2) Orienting regulation to encourage a level of develop- automation of final usages of electricity (including “do- ment of “smart” applications that can minimise the motics” in the widest context of the “Internet of Things”). cost borne by the regulated systems and, ultimately, end users, particularly the cost of communication ser- The blurring edge between the electrotechnical layer and vices that sometimes account for the lion’s share; the ICT layer in digitalised power systems implies a key regulatory challenge. So far, regulation has been tradi- 3) Ensuring that the widespread adoption of M2M appli- tionally thought to be based on a vertical (single-sector) cations (encouraged in part by the regulator) in the jurisdiction and, in most countries, organised with secto- energy and water networks does not pose an obstacle rial bodies. Only in few countries (such as Germany) the to the development of multisector solutions, (for ex- regulatory authority is a cross-sector body. In such an ample, smart cities based on shared communications institutional context, cooperation between energy regu- infrastructure). lators and telecom regulators is a natural byproduct of Moreover, particular attention has been drawn on the digitilisation (BEREC, 2017). latency parameter of M2M telecommunication services, in An example of cooperation between sectorial regula- terms of monitoring, control and protection. tors is ongoing in Italy. In 2013, the telecom regulators

105 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

3.3.1 Short-term measures to mitigate of network companies. Without knowing how to the impact of high levels of estimate this impact, revenues could be insuffi- 3 distributed generation on cient to cover distribution costs – and distribu- distribution revenues tion companies may incur financial losses. Or the opposite could happen, and network users could end up paying too much. Whilst these deviations Distribution revenues should be independent can be corrected ex post, retroactive actions of the volume of energy distributed. should be avoided because they tend to create regulatory uncertainty and jeopardise the regu- The combination of conventional regulation and lator’s credibility. Moreover, corrections are often the high penetration of DG may result in rising hard to apply in practice given existing informa- distribution costs and decreasing revenues for tion asymmetries, e.g., regulators would need to distribution companies. This impact can be coun- determine ex post whether avoided investments teracted in the short term through relatively sim- in distribution are due to decreased demand or ple regulatory modifications. due to efficiency gains. Various ways to address these issues are discussed in Section 3.3.2. One of the most immediate consequences of DG, when connected behind the customer’s meter, is Unfortunately, the conventional tools used by reg- that the amount of energy that is metered de- ulators to assess distribution costs are ill-adapted creases. Additionally, as discussed above, net- to the above-mentioned task. In a review of in- work investments may even need to increase ternational practices, Jamasb and Pollitt (2001) due to the impact of DG. This is because network find that the most common benchmarking meth- costs typically do not directly depend on the vol- ods rely on past information and still consider the ume of energy distributed. Moreover, network amount of energy distributed as one of the main operators have little control over this factor, thus cost drivers. It is urgent that, as discussed in Sec- being largely exposed to volume risks. Therefore, tion 3.3.2, these limitations are overcome through in case of high level of DG, remuneration should the adoption of new regulatory tools able to cap- be independent of the amount of energy distrib- ture the effect of DG and advanced grid technol- uted in order to prevent hurting the distribution ogies so that regulators may compensate distri- utilities financially.10 This separation is usually re- bution companies for the incremental DG-driven ferred to as revenue decoupling. costs and that these companies do not oppose the connection of DG: The key element of a revenue-decoupled remuneration, as compared to a conventional volu- “We recognise that there is significant uncer- metric remuneration, is that electricity rates are tainty around the volume of DG that will con- adjusted to ensure that the utility recovers the nect in DPCR5, its generation type, location amount of revenues initially determined by the and voltage, all of which make it very difficult regulator. A study of the implementation of reve- to anticipate the cost of connecting the DG to nue-decoupling mechanisms in US utilities shows the networks. We want the DPCR5 DG incentive that, by the end of 2012, 15 states had implement- to ensure that there is still a strong incentive on ed some form of revenue decoupling for at least DNOs to connect DG, and protect them from one of their electric utilities, comprising up to 24 the risk of increased connection costs DG”. electric utilities (Morgan, 2013). (OFGEM, 2009a: 22) Revenue decoupling effectively addresses the is- Acknowledging this problem, and as a transito- sue of revenues being depressed by DG. Never- ry measure, the United Kingdom’s energy regu- theless, regulators will still need to pass the im- lator implemented a specific DG incentive for this pact of DG growth on to the allowed revenues purpose. This incentive was in place for two con-

10. Similar problems may arise in the short term as a consequence of energy-efficiency measures or any other cause of falling demand. If any of these measures reduce also reduces peak load, distribution costs may indeed decrease over the long term.

106 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

secutive regulatory periods, between 2000 and be adapted to encourage distribution utilities to 2010, before the implementation of the RIIO re- adopt innovative approaches towards grid plan- form (see Box 3.7). This incentive comprised two ning and operation (CEER, 2014). mechanisms which provided distribution network In fact, several regulatory commissions have car- operators with an additional remuneration de- ried out (or are carrying out) various forms of pending on the amount of DG connected. public consultation and review processes to eval- The estimated costs incurred by distribution uate how future regulatory frameworks should companies to connect DG were partially passed look and the main reforms needed to achieve key through to tariffs and subject to an accelerated goals. Salient examples include those of the United depreciation. More specifically, companies could Kingdom (OFGEM, 2010b), Italy (AEEGSI, 2015a) recoup 80% of the related investments over 15 and the state of New York (New York DPS, 2014). years (the depreciation period considered for These are described in more detail in Box 3.7. other assets was 20 years). Companies were also given an additional amount, between £1 and £2 per every kW of DG capacity connected over an Regulatory reforms to encourage the additional 15 years. Box 3.93.7 adoption of innovative grid technolo- Combined, these two regulatory schemes might gies and solutions neutralise the negative effects DG could have on DSO revenues. However, they may still be not United Kingdom: The pioneering RIIO reform sufficient to achieve the degree of innovation Starting point: The United Kingdom was a forerun- and transformation in the distribution networks ner in the application of Retail Price Index (RPI)-X which, as discussed in Section 3.2, is necessary to regulation to the energy network industries. Over achieve an efficient integration of high DG. time, distribution regulation evolved towards greater separation of revenues from the amount of ener- 3.3.2 Guidelines for smarter gy distributed, a more prominent role of compara- distribution regulation tive benchmarking across firms, stronger emphasis on the quality of service provision and equalisation of incentives to reduce both capital expenditures Conventional, investment-focused regulation (CAPEX) and operational expenditures (OPEX) (OF- should evolve to match the new roles of GEM, 2009b). In fact, the regulation during the last distribution companies regarding DER regulatory period in which the RPI-X approach was integration. formally applied (2010–15) already contained many Remuneration to distribution companies of the elements that characterised the subsequent should have a long term view. Flexible reform.

remuneration schemes should be adopted. Regulatory reform: In 2009, the energy regulator – Economic incentives to distribution companies the Office of Gas and Electricity Markets (OFGEM) should focus on both operational and capital – launched a comprehensive review of network reg- cost. Forward-looking cost assessments and ulation, which was called RPI-X@20. While acknowl- long-term detailed utility investment plans edging that RPI-X regulation had worked well, the should be used. report asked whether new regulatory approaches were necessary for the challenges ahead (OFGEM, 2009c). This process resulted in a new regulatory Pilot projects are an essential step to testing model, called RIIO for “setting Revenue using Incen- technological solutions and disseminating knowl- tives to deliver Innovation and Outputs” (OFGEM, edge gained and best practices. However, the 2010b). long-term transformation of the distribution grid The major features of the RIIO reform were as fol- cannot rely exclusively on subsidised experimen- lows. Regulatory periods were increased from five tation. Over the long term, regulation needs to

107 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

to eight years, ex ante total expenditure (TOTEX) revenue to eight years. Input-based innovation incentives will re- allowances were based on well-justified business plans main for non-tested smart grid solutions, with a particular 3 and a comprehensive cost assessment toolbox, the num- focus on the modernisation of low-voltage networks and ber of incentives based on output factors (customer sat- the development of second-generation smart metering. isfaction, environmental impact or energy not injected by New York: Proposals for Reforming renewable energy sources due to network unavailability) the Energy Vision – REV was increased, and automatic revenue was adjusted within the regulatory period (menu regulation, profit-sharing Starting point: The traditional utility regulation mod- and reopeners) (OFGEM, 2010a). The RIIO model is being el in New York was based on the conventional cost of first applied to power distribution in the period 2015–23. service or rate of return regulation. Nonetheless, rate cases progressively evolved towards multiyear reviews Italy: Transitioning from input-based that implicitly provided utilities with incentives to cut demonstration to output-based deployment costs. Therefore, earning-sharing mechanisms and per- Starting point: The Italian approach to distribution regula- formance-based schemes, mainly related to quality of tion was based on the combination of an incentive-based service, were also introduced in order to protect both revenue cap on OPEX and a more conventional rate-of- consumers and utilities from potential deviations with re- return regulation for CAPEX. In 2010, the Italian regulator spect to the conditions initially considered by the regula- AEEGSI implemented an incentive scheme to promote tor (New York DPS, 2014). smart grid projects with a focus on DG integration in the Regulatory reform: The review acknowledges that con- medium-voltage network. Under this scheme, smart grid ventional regulation will not drive the desired evolution investments approved by the regulator were awarded an towards a distribution system platform provider mod- additional 2% in their allowed rate of return for a period of el. The proposals released so far draw heavily from the 12 years (9% pretax in total) (CEER, 2011). United Kingdom’s experience: longer regulatory periods, Regulatory reform: AEEGSI has recently announced its a shift towards an output-based remuneration (both for willingness to build on the experience collected from revenue adjustment and for monitoring only or score- the previous demonstration projects and move “from in- cards), use of flexibility mechanisms such as reopeners put-based demonstration to output-based deployment” and earning sharing schemes and the encouragement of (Lo Schiavo, 2015). A public consultation process was an efficient allocation of operational and capital expen- launched in May 2015 (AEEGSI, 2015a). The proposals ditures. made by the regulator in a resulting report offer two tar- In this case, a TOTEX-based regulation to mitigate the gets: equalising incentives to reduce OPEX and CAPEX CAPEX bias would be hindered by US accounting poli- (i.e., shifting towards a TOTEX regulation) and promoting cies.a Another factor that sets the REV reform apart from the deployment of smart grid functionalities presenting its European counterparts is that, under the distribution positive benefit-cost ratios through technical perfor- system platform provider model, utilities are encouraged mance indicators. This document was followed by a sec- to seek additional revenue streams beyond existing rate- ond round of consultations in September 2015 (AEEGSI, based earnings in the form of value-added services pro- 2015b) and a document that identified two main function- vided to other stakeholders (New York DPS, 2015). Ob- alities to be covered by output-based regulation in the taining these so-called market-based earnings may not upcoming regulatory period, i.e., medium-voltage (MV) be possible for European distribution companies due to grid observability and voltage control in MV networks. existing rules on the unbundling of activities. Furthermore, it states that the power of the incentive mechanisms depends on the degree of their implemen- a. The US Generally Accepted Accounting Principles and New tation and the technical capabilities of the solutions im- York regulations state that utilities are entitled to recover plemented. assets based on original cost less depreciation as opposed to a regulatory asset base determined by the regulator (and which The regulator has recently implemented provisions, in- is not necessarily based on original costs). US utilities’ adoption cluding incentives for the implementation of smart grid of the TOTEX regulation under the aforementioned accounting rules could force them to write off certain assets since it is not solutions in areas with high penetration of DG, a bo- straightforward to demonstrate that a specific asset is being nus-malus scheme related to the modernisation of old recovered through the rates (New York DPS, 2015). urban grids, and the lengthening of regulatory periods up

108 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Despite their different scopes and approaches, crease transparency, provide information rele- several common trends and guidelines can be vant to system planning and prepare future ad- found among the most advanced proposals for ditional incentives schemes (New York DPS, regulatory reform. These are discussed below 2014). Among the indicators proposed are asset and summarised in Table 3.2. utilisation metrics, DG penetration, emissions reduction, customer satisfaction and EV adoption Output orientation: regulation ought to shift the (New York DPS, 2015). focus of regulatory scrutiny from investment adequacy (inputs) to how well distribution com- Efficiency incentives neutral to the actual cost panies are providing the required services to structure (CAPEX versus OPEX): regulation network users (outputs). Thus, distribution com- should encourage distribution companies to re- panies should be increasingly assessed on the alise the contribution of DER to distribution net- basis of a set of output indicators beyond the work planning and the benefits of smarter grids, common ones related to supply interruptions and which implies exploiting the trade-offs between energy losses. These indicators can be used for CAPEX and OPEX, e.g., investment deferral. In- monitoring, for the purposes of comparison alone stead, existing regulatory approaches tend to en- or even to affect distribution revenues (depend- courage companies to resort to investment-based ing on performance levels). solutions. For instance, New York authorities have recent- In order to rectify this, the Office of Gas and Elec- ly proposed to relate distribution revenues to in- tricity Markets (OFGEM) has implemented two dicators, such as peak load reduction, customer mechanisms in the latest application of RIIO to information or time taken to process connection electricity distribution. First, cost assessments are requests. Additionally, monitoring other types performed following a total expenditure (TOTEX) of indicators has been proposed as a way to in- approach, i.e., the joint consideration of all com-

Table 3.2 General guidelines for a smarter distribution regulation

Shift towards an output-oriented regulation

Conventional, investment-focused regulation should evolve to match the new roles of distribution firms regarding DER connection, environmental impact, customer satisfaction and social obligations. Thus, additional output indicators should be used for performance monitoring or as utility revenue drivers.

Remuneration formulas neutral to the actual cost structure (CAPEX vs OPEX)

Cost reduction efforts tend to focus on OPEX due to an input orientation and short time intervals between price reviews. However, exploiting smart grid benefits such as investment deferral thanks to a more-active network management and DER requires equalising the incentives to cut both types of expenditures.

Forward-looking cost assessment

Due to rapid technology change and DER penetration, revenue allowances may no longer be based exclusively on past expenditures and benchmarking studies relying on historical information. Forward-looking cost assessments and long-term detailed utility investment plans should be increasingly used by regulators.

Price reviews with a long-term view

Shifting the goal of regulation from investment adequacy to long-term efficiency raises the question of extending the length of regulatory periods, with intermediate monitoring and review processes to check for excessive deviations.

More flexible remuneration formulas

Utilities are bound to face higher uncertainties driven by technology change and uncertain DER connection. This, together with potentially longer regulatory periods, strengthens the need for more flexible remuneration schemes such as profit- or earning-sharing mechanisms or index-triggered reopeners.

109 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

panies’ expenditures regardless of type (OFGEM, In this context, New York is planning to extend 2014). Second, a fixed percentage of the TOTEX is its multiyear rate plans from three to five years, 3 included in the regulatory asset base and subject whereas the United Kingdom has already extend- to a recovery period of 45 years. Thus, the remu- ed the frequency of price controls from five to neration is independent of the actual cost struc- eight years. The flip side of this measure is that ture of the companies (OFGEM, 2013b). uncertainties increase. Flexibility mechanisms and automatic revisions are then usually needed.11 Forward-looking cost assessment: regulators ought to adapt their cost assessment tools and Flexible remuneration formulas: as the distribu- their application, as well as request distribution tion networks evolve, technology and demand for companies to provide them with detailed invest- network services will no longer be easily predict- ment plans. The goal is to capture the real condi- able. If the time between price reviews is length- tions that will be faced by the distribution compa- ened, the uncertainties are worsened. Therefore, nies and avoid long lags between the time costs remuneration formulas should be flexible in order are incurred and when they are recovered. to react to unforeseen events and deviations. In practice, the main tools for this are the so-called The application of the RIIO approach to electric- profit- or earning-sharing mechanisms that pre- ity distribution in the United Kingdom provides vent distribution companies’ excessive loss or a good example of using a forward-looking ap- gain due to the incorrect setting of baseline rev- proach in cost assessments. For instance, despite enues. When these deviations exceed a certain applying conventional regression benchmarking level, the baseline should be revised by means models, a combination of past information and of reopeners. This implies that if actual costs de- forecasted data is used as input for the models viate more than a certain amount from baseline (OFGEM, 2013c). Moreover, distribution compa- revenues, the regulator is entitled to intervene. nies were mandated to submit detailed business plans. As part of this process, the companies had The use of earnings-sharing mechanisms is a stan- to apply an engineering model, called the Trans- dard in New York. The regulator monitors the reform model, to justify that their investment plans turn on equity of utilities. When this exceeds a pre- captured the effects of smart grids and DG (OF- defined threshold, the benefits are shared between GEM, 2013a). the utility and the ratepayers. The innovation being proposed is to determine the sharing factor to the Focus on long-term efficiency: the principle of performance indicators (outputs) of the utilities. encouraging long-term efficiency over myopic or Thus, utilities presenting a better performance or strategic short-term cost reductions is embedded scorecard will be allowed to retain much more of in all the aforementioned guidelines. Regulators the efficiency gain as compared to poor-perform- should progressively extend the length of regu- ing utilities. OFGEM has implemented a series of latory periods to force distribution companies to reopeners in case different types of expenses de- adopt this long-term viewpoint. This would al- viate significantly from the forecasts, as well as an low distribution companies to spend their efforts intermediate review of the outputs delivered by in improving their performance rather than on distribution companies (OFGEM, 2013d). time-consuming rate cases.

11. Jenkins and Pérez-Arriaga (2014) present an advanced implementation proposal of this guideline. The authors build on the menu regulation with profit-sharing contracts applied in the United Kingdom by introducing an automatic ex post review based on adjustment factors computed in advance. These factors automatically correct revenue allowances for deviations in the load growth and DG penetration, thus mitigating the risks derived from forecast errors. Automatic revisions may be hampered by acceptability issues on behalf of consumer advocates (which usually play an active role in US ratemaking), particularly when they can lead to tariff hikes. A noteworthy case is that of ComEd in Illinois in 2011. A bill was introduced enabling this utility to invest in grid modernisation and smart metering in exchange for automatic rate increases that granted the utility a return on equity of 10%. This bill was opposed by consumer advocates and the state attorney general, and it was even vetoed by the state governor. It was finally passed with a “veto-proof” majority.

110 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

3.3.3 Conclusions •• Distribution companies should not be incenti- and recommendations vised to invest in conventional network assets The efficient integration of DG on a large scale when a less-costly alternative involving higher is not feasible without appropriate adaptation of operation expenditures exists. Therefore, con- regulatory practices. The focus so far has been trary to widespread practices, incentives for placed mainly on neutralising the potential nega- cost reductions should be placed on TOTEX. tive effect that DG may have on the revenues and •• Cost-assessment methodologies should in- investment needs of distribution companies. creasingly rely on forecasted data and well-justified investment plans submitted by the regu- On the one hand, distribution revenues should lated companies. be made independent of the volume of energy •• Promoting efficient investment in distribution distributed. Otherwise, self-consumption or ener- networks requires adopting a long-term per- gy-efficient measures may immediately translate spective, given the long useful life of assets. In into a reduction in revenues, whilst costs remain practical terms, this has motivated some reg- constant or even increase. The main component ulators to lengthen the period between price of revenue decoupling consists of an ex post reviews. adjustment of network tariffs so that the utili- •• Remuneration formulas should be more flexible ty exactly recoups the revenues allowed by the to account for this. Deviations between actu- regulator. al costs and revenues allowed ex ante may be On the other hand, as long as cost-assessment subject to profit-sharing mechanisms and even regulatory tools are unable to capture the impact lead to a revision of the allowances in case this of DG on distribution costs, it is necessary to set deviation exceeds a certain threshold. economic compensations for distribution companies due to the cost increase driven by the connection of DG. These compensations should be added on top of conventional revenue allowances and be exempt from efficiency requirements. However, none of the aforementioned regulatory measures by themselves actually encourages distribution companies to innovate grid planning and operation, which is a must for efficient DG integration. Deeper changes to current regulatory approaches are required to achieve this transformation, as has been acknowledged by some regulatory authorities. Revenue decoupling is a first step, to be followed by the implementation of new regulatory frameworks that promote long-term efficiency in the presence of large volumes of DER. A review of the ongoing regulatory reforms of this nature reveals several guidelines for future reforms of regulatory frameworks for power distribution:

•• The focus of regulation ought to be shifted from ensuring that companies carry out enough network investments to assessing them on the basis of their performance, measured by an extended set of indicators.

111 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

3.4. SELF-CONSUMPTION: od (IEA, 2016), thus leading to inefficient invest- TARIFFS AND METERING ment decisions and consumption choices. 3 The challenge for policy makers is twofold: 1) Self-consumption can yield benefits both eliminate barriers to the adoption of self-con- for end users and for the power system as sumption, especially for commercial and residen- a whole. At the same time, it can promote tial customers and 2) revisit tariff designs for net- demand side flexibility, including the adoption work and other regulated charges adopting more of distributed storage behind the meter. advanced retail energy pricing mechanisms that Policy and regulation should actively promote appropriately value both the energy self-con- self-consumption and remove administrative sumed and the energy exported to the grid. barriers. Consumer empowerment and the diffusion of DG In order to attain a sustainable development have paved the way for end-consumers who lo- of active agents with consumption, production cally generate part of their electricity needs and and storage, it is important to adopt a cost- who, during some periods, inject the surplus ener- reflective design of retail tariffs and support gy into the grid (thus becoming a producer/con- the roll out advanced metering technologies. sumer). Local self-consumption can deliver savings both to the end-consumer and the system as a whole (although, as discussed below, these Self-consumption12 is a RES promotion policy with two do not necessarily go hand in hand). Box 3.8 a relatively simple design and with high levels of reports two examples of commercial consumers consumer acceptance. Despite the benefits to that have managed to reduce their electricity bills consumers and overall system efficiency, in case to a significant extent thanks to self-consumption of high DG levels, conventional tariff designs and with solar PV. net-metering policies13 may jeopardise a system’s cost-recovery and create cross-subsidisation From a system perspective, the benefits of among those customers who self-consume and self-consumption are realised when local gen- those who do not. eration coincides in time with the consumption behind the meter. Under these conditions, self- It is important to highlight that end-consumers consumption reduces the utilisation rate of net- make their decisions based not on wholesale work assets, both at the transmission and distri- electricity prices, but on the retail tariffs they bution level. This reduces power flows through pay. In fact, the cost of generating electricity of- the grid, decreasing energy losses, especially in ten accounts for less than half of the final electric- the distribution network whose losses account ity costs paid by the end-user. Another relevant for most of total system energy losses, and re- clarification is related to the concept of grid par- ducing peak demand, thus potentially postpon- ity, i.e. the point at which the LCOE of (usually) ing the need for network reinforcements and up- PV generation falls below the volumetric charge grades in areas presenting scarcity of spare grid (USD/ kWh). The adoption of PV panels on con- capacity. Note that the benefits of investment de- sumer premises might be spurred once this break- ferral may be realised when the distribution com- even point is reached. But this may not happen pany has some certainty about the reduction of when the retail tariffs do not reflect the actual peak net demand, e.g., by having the possibility value of electricity at each location and time peri-

12. Self-consumption policies aim to promote the use of on-site generation for local consumption. The part of the bill that can be compensated varies per different approaches. A ‘self-consumption scheme’ usually refers to the mechanism of energy consumption in real-time (e.g. per 15 minutes). Schemes that allow compensating electricity injection into the grid and electricity consumption during a larger timeframe are usually called ‘net-metering schemes’. Schemes that allow the calculation of the compensation on a cash-flow basis, rather than on an energy basis, are usually called ‘net-billing schemes’. Hybrid schemes also exist (IEA-PVPS, 2016). See also Box 3.9. 13. See also ‘Review and analysis of self-consumption policies’, IEA-PVPS, 2016

112 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

to manage the injection/withdrawal of network Box 3.8 Examples of self-consumption benefits users. This topic is discussed in further detail in for commercial end users Sections 3.2.2 and 3.5.1. The European Commission (2015c) reports the case How to remunerate this local production is a key of a commercial consumer in Germany, further de- policy decision. Under net-metering schemes, this scribed in Kraftwerk (2015), which is capable of production is valued equal to the energy generat- self-consuming 87% of the annual production of its ed for self-consumption, i.e., at the retail price (as 63 kWp photovoltaic (PV) plant.a This is achieved discussed in Section 3.4.4). Some legislation, like because the manufacturing process runs mainly that recently passed in Spain (MINETUR, 2015c), throughout the day, as shown in Figure 3.9. On an eliminates the incentive to the surplus, giving it a annual basis, the company is able to decrease its an- value of zero. Another option is to link the value nual electricity bill by more than 50 000 kWh, which of that generation to the wholesale energy price. amounts to over 15% of its annual consumption.

Several good practices have been shown to pro- Another illustrative case is an Australian winery that mote the adoption of self-consumption by install- installed a 90 kWp solar rooftop.b The solar system ing DG behind the meter and to incentivise end is estimated to reduce the winery’s carbon emissions users to not only install DG, but also to align their by 22% and results in annual savings of up to AU$ 26 production with their consumption profile (by in- 000 (approximately USD 18 900). creasing the amount of self-consumed energy).

Policy recommendations identified by the Euro- Figure 3.9 Sample daily load profile for a plastic pean Commission (European Commission, 2015c) manufacturing company with self-consumption include the following: from PV •• Allow the installation of DG and storage behind Sample daily proﬁle the meter, even for small commercial and resi- Generated dential consumers. Operational solar enegy •• Simplify authorisation procedures for this kind energy consumption of installation, allowing simple notification to Grid supply the distribution grid operator. •• Promote the deployment of smart meters and KW allow aggregators that facilitate the participa- Feed-in to grid tion of end users into wholesale and retail mar- Self-consumed kets. solar energy •• By transmitting the right electricity prices to end users, e.g., dynamic hourly prices, promote 0 2 4 6 8 10 12 14 16 18 20 22 24 their reaction through demand-side response Hours of the day actions and installation of distributed storage.

Source: Kraftwerk, 2015 3.4.1 Net-metering policies Net metering constitutes a step beyond self-con- a. kWp measures the maximum power (in kW) instantly deliverable from a PV. sumption, and consists of allowing prosumers to b. Detailed information on this case can be found at https:// use their production surplus in a given period to www.q-cells.com/consumer/commercial-and-industrial/ref- compensate for their consumption in a different erences.html#references_1046894. moment. Because of this, net-metering is sometimes characterised as allowing end users to use the distribution grid as storage for their energy surplus. In fact, the implications of net-metering are much deeper. These are mainly related to the fact that, under net metering, the energy

113 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

injected into the grid is implicitly valued at the re- Box 3.9 Key terms related to self-consumption tail electricity price. Detailed, relevant definitions and net-metering 3 and concepts are provided in Box 3.9. −− Billing period: corresponds to how often end-con- As shown in Figure 3.10, net-metering is a wide- sumers receive and pay their electricity bills. This spread policy in the United States, present as period typically ranges from one to a few months. a mandatory rule in 41 states and several other jurisdictions. According to SEPA (2015), 99% of −− Metering interval: interval of time for which the me- solar installations across the United States were ter records (net) consumption. In the case of modern electronic meters, this can be up to 15 minutes, under a net-metering scheme in the year 2014, whereas in the case of conventional electro-me- representing 44% of total solar PV installed ca- chanical meters, this would correspond to the peri- pacity in the nation. For the sake of comparison, od in which the period meter reading is taken and, it is worth mentioning that, by contrast, feed-in normally, also the billing period. tariffs (FiTs) are present in seven states, and the Renewable Portfolio Standard (RPS) are present −− Netting period: time interval during which prosum- in 29 states.14 ers are allowed to compensate energy injections and withdrawals, i.e., once this time expires, the pro- On the other hand, the deployment of renew- sumer is no longer entitled to compensate net con- able DG in Europe has been mainly driven by di- sumption with local excess production. This interval rect support schemes, such as FiTs or tradable depends on the design of the net-metering scheme green certificates (CEER, 2015c). Notwithstand- implemented by regulators, usually ranging from a ing, net-metering, net billing, and self-consump- single hour to a whole year.a

tion have been introduced in several Europe- (a) When local production is not controllable, and once the an countries (see Figure 3.11). Such schemes are DG installed capacity has been fixed, setting different seen by policy makers as a way to keep promot- netting periods does not affect physical power flows through the grid. In other words, if the consumption and ing the adoption of RES, mainly solar PV, whilst generation profiles are set, power flows are independent progressively phasing out existing support pay- of metering and billing arrangements. Nonetheless, economic transactions may differ in significant ways. ments. Moreover, self-consumption encourages

Figure 3.10 Net-metering policies across the United States in July, 2016

State-developed mandatory rules for certain utilities (41 states+DC+3 territories)

No statewide mandatory rules, but some utilities allow net metering (2 states)

Satewide distributed generation compensation rules other than net metering (4 states+1 territory)

41 States + DC, AS, USVI, & PR have mandatory net metering rules U.S. Territories: AS PR VI GU

Source: DSIRE, 2016

14. Data extracted from the Database of State Incentives for Renewables and Efficiency (DSIRE.®). Accessible at: www.dsireusa.org 114 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Figure 3.11 PV production in the European Union in 2014

% 5 3.5% 4

Italy Spain Greece Cyprus France Austria Germany Bulgaria Belgiuim Romania Slovenia Slovakia Denmark Portugal Total Europe The Kingdom Czech Republic United Kingdom of Netherlands PV penetration Self-consumption Net-biling Net-metering

Source: Solar Power Europe, 2015 locating DG close to where electricity is actually this practice can jeopardise a system’s cost-re- consumed, thus improving network utilisation. covery rates if tariffs are not cost reflective and if conventional metering technologies with very In fact, the European Commission’s initiative limited capabilities are used. “Clean Energy for All Europeans” is proposing to mandate that EU Member States enable renew- Under self-consumption, the value of self-con- able self-consumption, including the right to sell sumed renewable electricity seen by end users is excess production, and prevent “disproportion- determined by the level of the retail tariff, more ately burdensome procedures and charges that specifically the energy term (USD/kWh). Howev- are not cost reflective” (European Commission, er, current retail tariffs do not necessarily reflect 2016b). Self-consumption has been considered the actual cost of electricity at a given moment. It so relevant that this provision has been included is acknowledged that the value of electrical ener- in two different directive proposals, i.e., the elec- gy depends on the times when it is produced or tricity Directive and the RES Directive (European consumed. Electricity markets value the energy in Commission, 2016b, 2016c). different time frames, ranging from hourly prices to real-time prices in periods of several minutes. 3.4.2 Self-consumption and the missing Generally, this price volatility is not transmitted money problem to retail prices. Frequently, retail prices adopt the form of flat rates or, in some cases, energy time- Self-consumption mechanisms with hourly or of-use rates. even shorter netting intervals would contribute What is more, energy-based rates are often in- to the sustainable development of on-site tended to recoup system costs that do not de- generation. pend on the amount of energy consumed, such as network costs, system operation costs, costs of Self-consumption involves a straightforward im- efficiency programs or support for domestic fu- plementation process, high levels of customer els or RES. Therefore, the amount of the system’s acceptance, and low transaction costs. However, fixed costs that end-users with self-consumption

115 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

installations, especially those under net-meter- Furthermore, regulators may be forced to raise ing schemes, defray can fall if tariffs do not re- retail tariffs to ensure cost-recovery. Higher rates 3 flect costs. Consequently, when self-consumption also increase the economic incentive to become occurs under conventional tariff designs – with a self-producer, creating a vicious circle. Also, mostly a volumetric, i.e., USD/kWh, component the potential large-scale adoption of distributed – consumers without DG are de facto subsidising storage can result in prosumers disconnecting the ones who adopt such practices. Net-meter- from the main grid (see figure 3.12) – a process ing schemes intensify this effect and reduce the that some have called the utility death spiral (EEI, sensitivity of end users to variations in electricity 2013; Paulos, 2015). costs at different times, undermining efforts to Although grid defection is a real concern of many promote demand-response and to promote high utilities, this possibility should not be overstated; RE feed-in when it is critical to do so from a sys- not all users have the financial resources, willing- tem-wide perspective (CEER, 2016). ness and space to disconnect from the main grid. It should be noted that DG can bring significant Understanding how many users do would require benefits to the power system: it decreases pro- specific and comprehensive analysis of each ju- duction costs and losses and defers investment risdiction (see RMI, 2014, for an example), a proof new capacity. However, net-metering may not cess that goes beyond evaluating the economics the most suitable way to compensate DER own- of solar-plus-battery systems as compared to re- ers. Instead, enabling self-consumption with a tail tariffs. EPRI (2016) suggests that the cost and cost-reflective pricing system, as described in reliability implications of disconnecting from the Section 3.4.4, would benefit DER users when they main grid are often underestimated. In any case, have a positive impact on the system. Cost-re- decreasing utility revenues and cross-subsidisa- flective tariffs would promote efficient DER in- tion are growing concerns for policy makers and vestments, bringing additional value to the sys- are being tackled through different policy alter- tem as a whole as a whole. natives.

Figure 3.12 Possible effects of traditional regulatory approach in the presence of high shares of DER

End-user Perspective: · Technology cost reduction (PV, storage) · Increased consumer awareness and information

· Tari increases · Increased DER · Stronger incentives · Reduced revenues for DER adoption · Grid defection

Utility Perspective: · Fixed costs not fully recovered · DER connection costs · Grid modernisation investments

116 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

3.4.3 Short term solutions to tackle duced. For instance, although The Kingdom of the missing money problem Netherlands simplemented a yearly net-metering In the absence of a well-established methodolo- scheme, prosumers only receive compensation gy to calculate truly cost-reflective tariffs, regula- of up to 5 000 kWh/year. Similarly, Denmark re- tors may adopt ad hoc rules to limit the amount cently changed the netting period from one year of missing money created through self-consump- to one hour, making it a purely self-consumption tion and potentially worsened by net-metering scheme. policies when a high share of DG is installed. 3.4.4 Towards a new tariff To prevent prosumers from oversizing their gen- design approach eration units, some limitations have been set on the allowed installed capacities per metering Moving away from purely volumetric point. Most net-metering policies in the United charges towards cost-reflective tariff States also include cumulative aggregate caps. structures is required. For example, PV systems in New York state may not surpass 25 kW for residential consumers, 100 Cost-reflective network charges should kW for farms and 2 MW for non-residential users, be allocated to end-users taking into whereas aggregate installed capacity is limited consideration their location, net hourly to 6% of the utility’s demand for the year 2005 consumption/injection and impact (including solar, farm-based biogas, fuel cells, on asset utilisation. micro-hydroelectric and residential micro-combined-heat-and-power). On the other hand, DG The introduction of a rigorous design for network units in California may go up to 1 MW in capac- and retail tariffs that sends efficient economic ity. An overall net-metering programme cap at signals to distribution network users is the ap- the utility level was defined by the Public Utilities propriate long-term solution to enable sustain- Commission, beyond which the investor-owned able self-consumption and encourage demand utilities (IOUs) are no longer obliged to offer response and storage. The following are some net-metering to consumers. As of March 2016, steps taken by regulators in this direction. approximately 79% of the capacity cap had been reached (see Table 3.3). Abandon purely volumetric charges Adopting more cost-reflective tariff struc- The volume of energy that is self-consumed and tures requires moving away from purely volu- paid for can also be limited to mitigate problems metric charges and introducing some kind of of cost-recovery. Either the maximum amount of fixed charge (e.g., USD/meter-month) or de- energy consumed could be limited or, indirect- mand charge (USD/kW). As highlighted in NREL ly, the length of the netting period could be re- (2015b), introducing these changes in rate structures involves careful consideration of the im- Table 3.3 Net-metering cap at the utility level in pacts they will have on consumer bills as well as California (March 2016) on the volumetric component. In addition to addressing the missing money problem, capacity Remaining charges encourage prosumers to reduce stress 5% NEM Capacity Utility Capacity on the distribution grid. In other words, the: March 2016 [MW] [MW] “rate design for mass-market customers should begin to place a greater weight on the peak de- PG&E 2409 435.4 mand of the customer, which is closely related SCE 2240 643.7 to the cost of the system and which can be man- SDG&E 607 34.1 aged by the customer” (New York DPS, 2015:11).

Source: CPUC, 2016

117 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

In the same line, the State of California has recently cost-reflective tariffs, whereas a peak demand launched a proceeding which, among other goals, charge encourages consumers to install storage 3 aims to “ensure that the successor tariff is based on capacity for peak shaving, thus mitigating the im- the costs and benefits of the renewable electrical pact of PV penetration on the system. generation facility”.15 The residential rate structure Expose prosumers to time-dependent pricing was revised, after which the state’s Public Utilities Commission recommended utilities to implement The energy component of the retail tariff deter- either minimum bills or fixed charges on residen- mines the value of prosumers’ self-consumed tial consumers (CPUC, 2014). energy. Flat energy rates have been conventionally applied, especially to residential consumers, Box 3.10 shows how different pricing schemes can mostly for reasons of simplicity, limited metering affect end-users’ investment decisions. Net-me- capabilities and an alleged lack of demand flex- tering promotes the installation of higher PV ca- ibility. However, these assumptions are being in- pacities, probably above the optimal size under creasingly challenged by the ongoing changes

Box 3.10 Simulation of investment decisions in PV and storage under different pricing schemes

Figure 3.13 shows the results of an analysis aiming to purely volumetric charges; in the second, half of the net- estimate the effect of net-metering together with peak work costs are recovered through a peak demand charge. demand charges on end-users’ investment decisions It can be observed that net-metering significantly encour- regarding photovoltaic (PV) and storage systems un- ages PV investments, even in the presence of a demand der three different technology costs (Burger, 2015). The charge. Nonetheless, the break-even point for PV invest- results correspond to an average New York residential ments is affected by this demand charge. Investments house located in a suburb north of New York City with in storage, which in this case were not profitable with a a load profile based on information available at the US purely volumetric tariff, are also encouraged to meet the Department of Energy and Energy Information Adminis- evening peak demand. Storage may even be profitable tration. Two pricing scenarios for a monthly netting peri- for end-users in the absence of PV, thanks to its potential od are considered: in the first, consumers are exposed to for reducing peak demand.

Figure 3.13 Influence of pricing signals on PV and storage investment decisions

6 6 5 5 4 4 3 3 2 2 PV Installed (kW) Installed PV 1 1 Storage Installed (kWh) Installed Storage 0 0 PV Storage PV Storage PV Storage PV Storage With net-metering Without net metering With net-metering Without net metering

Without peak demand charge With peak demand charge

PV (1,750 $/kW) / Storage (250 $/kWh) PV (3,000 $/kW) / Storage (425 $/kWh)

PV (2,250 $/kW) / Storage (355 $/kWh)

Source: Based on Burger, 2015

15. http://www.cpuc.ca.gov/PUC/energy/DistGen/NEMWorkShop04232014.htm.

118 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

in the power sector. Truly cost-reflective elec- based rate design is likely to reduce the current tricity rates should include time differentiation. level of support provided to NEM […]. The goal The most straightforward form of time-depen- of promoting customer-sited DG is important, dent pricing would be time-of-use (ToU) tariffs, but […] subsidies should be explicit and trans- whereas the most advanced approach is based parent” (CPUC, 2014: 24). on hourly prices enabled by advanced metering “Best practice include: […] Preference for infrastructure (AMI) and, in liberalised settings, self-consumption schemes over net-metering by liquid and transparent wholesale markets. schemes […] Phasing in of short-term market The European Commission (2015c) provides rec- exposure by valuing surplus electricity injected ommendations on the implementation of new into the grid at the wholesale market price” (Eu- grid tariff designs to tackle the missing mon- ropean Commission, 2015c: 12). ey problem and to incentivise system efficiency In fact, several alternative approaches to net-me- through self-consumption. These include: tering, especially relevant in presence of high •• Future tariff designs should be based on objec- shares of DG, can compensate excess on-site DG tive and non-discriminatory criteria and reflect production. In Denmark, prosumers are paid a the impact of the consumer on the electricity FiT (0.08€/kWh) for their net electricity exports, grid while guaranteeing sufficient recovery of whereas in the United Kingdom, PV systems small- grid and system costs. er than 30 kilowatt peak (kWp) are given both a •• If modifications of the current tariff struc- generation tariff for the energy that is self-con- ture are deemed necessary and appropriate, sumed and a bonus for the excess electricity fed they should take into account the need to en- into the grid (European Commission, 2015c). sure stability for previous investments in self- As costs go down and technology matures, consumption installations. self-produced electricity will increasingly com- Decouple feed-in compensation from pete against centralised generation rather than retail tariffs against a retail tariff. In a recent Portuguese reg- Implicitly compensating the injection of electric- ulation on self-consumption, for example, the en- ity at the retail tariff level, as net-metering does, ergy injected into the grid is compensated at the is de facto a support instrument to RE produc- average value of spot market prices (reduced by tion that has demonstrated to be very effective 10% to account for grid costs). to promote the deployment of decentralised Figure 3.14 shows that different US states have RE solutions in several countries. However, this implemented mechanisms that decouple the val- mechanism may not adequately reflect the real ue of net excess production from retail tariffs. value of RE feed-ins at different moments. This is acknowledged by many policy makers: Advanced tariff designs as a driver for demand response and distributed storage “The current convention of crediting at the av- For end-users’ tariffs to reflect costs, tariff design erage retail rate may be either too little or too methodologies should acknowledge the different much […]. Through the calculation of the full added cost components of retail electricity rates. value of DER to the system the utility will be These components are: 1) the electricity price/ able to determine the total economic value of cost in each time period and location; 2) the cost the resource […] used as the basis of the credit” of the transmission and distribution networks and (New York DPS, 2015: 13). 3) other regulated costs, such as system opera- “Given the level of cross-subsidy that [net ener- tion or RES support.16 gy metering (NEM)] represents today, any cost-

16. Note that these components strictly correspond to the costs of electricity supply. Nonetheless, taxes and levies may amount to a significant share of the final retail tariffs, even exceeding 50% of the final price in some European countries, such as Germany and Denmark (IEA, 2016).

119 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

3 Figure 3.14 Treatment of prosumers’ excess energy across the United States in July, 2016 NEG credited at retail rate; credits do not expire

NEG credited at retail rate at ﬁrst, then credits expire or are reduced (e.g., to the avoided cost rate at the end of year)

NEG credited at less than retail rate (e.g., avoided cost rate)

NEG is not compensated

No statewide mandatory net metering rules

Source: DSIRE, 2016

Exposing end-users to energy prices that vary coup network costs and send end-users efficient over time and location within the grid would pro- long-term economic signals affecting location mote demand flexibility and distributed stor- and investment decisions. Cost-reflective net- age. Although this is done through ToU energy work charges should be based on the drivers for prices or hourly dynamic prices, more advanced network expansion (Bharatkumar, 2015), such as schemes based on distribution locational margin- location, net hourly profile (how much end users al prices (DLMP) have been proposed (Bohn et may be consuming or producing depending on al., 1984). DLMPs essentially extend the concept the hour) and users’ contribution to asset utilisa- of nodal prices applied at the transmission level to tion during the most critical periods of the year. distribution networks. Thus, DLMPs reflect elec- This methodological approach yields network tricity’s different values depending on where it is charges that are largely independent of the being consumed or injected in the network. This volume of energy injected or consumed, but are pricing methodology would therefore capture the rather based to a great extent on a fixed per- effect of energy losses (which mostly take place user component, which could be differentiated at the distribution level) and grid congestion. Dy- by voltage level or type of network user,17 and a namic energy prices provide end-users with effi- time-dependent capacity component (USD/kW) cient short-term economic signals that promote reflecting the impact of each end-user on the a rational utilisation of DG, storage and demand utilisation of network assets. The aforementioned response, either by themselves or in combination. capacity term would reflect the incremental Network costs, both transmission and distribu- network costs driven by the existence of network tion, are recovered through the so-called network constraints, either current or foreseen, in the charges. Properly designed network charges re- corresponding period.18 However, in the absence

17. This term could be calculated for each network user or group of consumers according to different criteria at the choice of the regulator: Ramsey, fairness, burden sharing between small and large consumers, etc. 18. Note that, if DLMPs were in place, which is not the case in any system, part of these congestion-driven costs would be embedded into these energy prices. However, DLMPs, as it happens with nodal pricing in transmission networks, would not be enough to recoup the full network costs.

120 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

of constraints, capacity charges may produce 3.4.5 The importance inefficient responses from network users to of metering arrangements reduce the loading of a grid that is not congested, The flexibility of regulators to implement their de- e.g., peak load reduction or storage installation. sired self-consumption/net-metering scheme can Thus, the remaining costs (sunk costs), which are be severely limited by insufficient metering capa- not driven by peak consumption or generation, bilities. Conventional electromechanical meters should be recovered through the fixed charge allow only simple schemes with no or scarce time to prevent the distortion of previous economic granularity and without compensation for net ex- signals. ported energy. Electromechanical bidirectional It is noteworthy that network tariffs ought to be meters, or a two-meter configuration, overcome technology neutral. This means that for each time this last barrier. However, the length of netting period and location the variable network charge periods would be limited by the meter-reading (USD/kWh) should be the same, but with the op- process. Since meters are read every one or two posite sign for consumption and injection. How- months, cost-recovery would hardly be mitigated ever, contrary to what is common today, these by reducing the netting period. should not depend on the type of user that is be- Therefore, electronic meters capable of recording hind the meter. Instead, for example, a power in- bidirectional energy flows every few minutes are jection from a generator should be valued equal a precondition for the most advanced net-meter- to that of a storage unit or a reduction in demand. ing schemes relying on time-varying tariffs and/ Lastly, retail tariffs include other regulated costs or very frequent netting periods. Should a new related to renewable energy support policies or meter installation or replacement be needed, social policies. These costs can be very relevant regulators should assess the suitability of setting in some countries and their allocation to end-us- the obligation to install modern metering devic- ers should not distort the short- and long-term es on prosumer premises. The implementation economic signals provided through energy prices of efficient self-consumption policies is not the and network charges. In this regard, several solu- only driver for smart metering. Section 3.5.2 will tions have been proposed. For example, these further discuss other issues related to smart me- costs could be removed from the electricity rates, tering, such as its role in enhancing consumer defrayed through the public budget or in the form awareness, retail market functioning and demand of a tax applied to all forms of energy consump- response. The section will also outline smart me- tion, not only electricity. Another option could be tering deployment policies as well as metering to recover these costs through a fixed charge per data management models. consumer, acknowledging the existence of different customer categories (residential, commercial, 3.4.6 Additional policy considerations industrial) and even subcategories (residential The implementation of the aforementioned pol- consumers with small, medium and large hous- icy measures may raise some additional consid- es, for instance), but independent of their energy erations that are not strictly related to self-con- and capacity injections of withdrawals. sumption and net-metering but are important to It is through the implementation of efficient short- take into account. and long-term economic signals (e.g., energy Regulators ought to bear in mind how changes prices and network charges, respectively) that in retail tariff structure would affect all consum- distribution network users will be encouraged to ers, not only new prosumers. For instance, intro- provide the full value of distributed storage and ducing a capacity charge would modify the eco- demand response. Moreover, when combined nomic signals seen by pre-existing prosumers, with DG, both resources can maximise self-con- thus modifying the conditions in which they eval- sumption and benefit the system as a whole. Ag- uated their investment in DG. Moreover, the in- gregators may act as mediators to exploit the ad- troduction of fixed or capacity charges can have ditional flexibilities offered by these resources.

121 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

unintended consequences for low-consumption network investment and enhancing consumer customers, who may see their bills suddenly in- engagement and participation. Self-consumption 3 crease. Therefore, gradual reforms and the pro- should therefore be supported by regulation. gressive implementation of tariff changes are Nonetheless, if not adequately designed, self-con- advisable (CPUC, 2014; European Commission, sumption measures may create a missing money 2015c; New York DPS, 2015). problem in traditional regulatory settings, that Burdensome administrative procedures and com- may jeopardise the financial viability of utilities. plex tariff options may pose a significant barrier The reason for this is usually a combination of to end-users, particularly residential consumers. a not adequate retail tariff design, which main- Another policy priority should be to reduce trans- ly relies on a volumetric term to recover all sys- action costs for prosumers through targeted in- tem costs, and old metering technologies, which formation, simplified administrative procedures are not capable of measuring bidirectional power or web-based calculators (CPUC, 2014; European flows with a high level of time differentiation. In Commission, 2015b). some contexts, the challenge may be worsened by high volumes of net-metering policies, which Another important issue to bear in mind is the implicitly value the energy injected into the grid impact that a potential massive deployment of at the retail electricity price. In response, regula- small-scale storage, coupled with DG, can have tors have tried to limit the allowed installed ca- on the power system, and distribution networks pacity or the amount of energy being compen- in particular. Prosumers, following price signals, sated, or shorten the period of time over which could use the storage systems to store electric- energy injections are reduced from the amount of ity and inject it back into the grid at different electricity billed and paid. moments in time. Potentially, this could make it even more difficult for system operators to These provisions, however, do not represent a forecast production from DG. This is yet anoth- sustainable long-term solution to the missing er argument in favour of implementing a set of money problem. The potential negative effects cost-reflective tariff structures. Moreover, DER of high level of capacity supported through net access to the upstream markets should be en- metering depends on the combination of mostly abled. This would require, among other things, volumetric (USD/kWh) tariffs and standard me- extending the conventional roles of distribution ters, which are applied and read monthly at the companies, strengthening co-operation between most. Therefore, it would be advisable to pro- transmission and distribution grid operators and mote self-consumption schemes with hourly or promoting the aggregation of DER. Section 3.5.1 even shorter netting intervals. In addition to this, discusses these issues in further detail. The key end-users, especially those with self-consump- regulatory challenge in this regard lies in how tion installations, should be exposed to well-de- to co-ordinate different economic signals when signed retail tariffs that reflect the value of ener- these enter into conflict, for example, when the gy at each moment as well as its impact on future provision of balancing services by a DG unit or investments. When additional incentives com- a prosumer results in congestion at the distribu- pensating surplus electricity that is not instan- tion level. taneously self-consumed are deemed necessary, these should be explicit. 3.4.7 Conclusions and recommendations Some policy makers and regulators have already started to implement provisions that could be Self-consumption through the installation of DG considered aligned with cost-reflectivity crite- and, in the future, storage behind the meter is a ria, such as introducing demand or fixed charges, policy alternative that benefits end-users by de- ToU tariffs or decoupling the value of electrici- creasing their electricity bills and benefits the ty feed-in from the level of retail prices. Truly system by reducing energy losses, encouraging cost-reflective tariffs would send efficient short-

122 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

and long-term signals to end users, reflecting costs, at least in part, through alternative rev- their contribution to system costs and ensuring enue sources such as public budgets or taxes. the recovery of total system costs. Their calcu- Well-designed retail rates, including adequate lation should consider the following components time granularity, can only be transferred to of system costs: 1) the price of the energy on the end-consumers if an advanced metering infra- market, with temporal and locational differentia- structure is available. Electronic meters capable tion; 2) the cost of the grid and 3) other regulat- of recording bidirectional energy flows every few ed costs. minutes are a precondition for the sustainable •• Cost-reflective prices of electricity acknowl- development of self-consumption and to spur edge its different value over time, and also by end users’ demand response, including distribut- location, reflecting network effects (congestion ed storage. and energy losses). Time-of-use or dynamic energy prices translated to end-consumers 3.5 FUTURE ROLE OF DISTRIBUTION with DG and future storage installations would COMPANIES provide efficient economic short-term signals. These signals would in turn promote the efficient operation of those resources, including Distribution companies have to bridge demand-response, generation dispatch and the gap between flexibility providers, markets storage management. and system operators. Moreover, they should •• Network charges should provide end-users integrate the flexibilities offered by DER into with economic signals promoting efficient loca- their planning and operational practices. tion and investment decisions, which should be Thus, distribution companies should based on the actual contribution of each net- adopt new roles as market facilitators work user to the network costs. Cost-reflective and distribution system operators. network charges should be calculated attending to the various drivers of grid expansion. Thus, they should be allocated to end-users taking The energy transition is undoubtedly affecting into consideration their location, their net hour- all segments of the electricity power industry. ly consumption/injection and their contribution Nonetheless, it is in the lower end of the supply to asset utilisation. The resulting charges would chain where these impacts are driving a deeper consist of a time-dependent capacity compo- transformation. The system’s decentralisation, nent reflecting the individual contribution to the deployment of ICTs and enhanced consum- network peak utilisation (both injection and er awareness are forcing distribution companies withdrawal), plus some fixed component relat- to reconsider their conventional roles as network ed to the grid connection and other costs that owners and operators. can be somehow shared by all network users. The agents located at the distribution level are •• The remaining regulated costs related to ener- gaining in importance and their active contri- gy or social policies, when transferred to elec- bution will become essential to ensure secure tricity rates, should not distort the previous system operation. In this context, distribution short- and long-term economic signals. Several companies represent stakeholders essential to solutions have been proposed for the alloca- connecting end-users with upstream markets tion of such costs – e.g., through a fixed charge and system operation. Moreover, utilities should or differentiated by customer size or category. integrate the flexibilities offered by DER into their Fixed charges can also have undesired effects, planning and operational practices by newly de- such as jeopardising low-demand consumers, fined mechanisms. Performing these new roles, debasing demand response and energy-effi- distribution companies will see themselves in- ciency efforts or encouraging grid defection. teracting more closely with other agents such as Policy makers may decide to recoup these suppliers, aggregators and TSOs/ISOs.

123 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Moreover, regulation should define the role to be tems. Distribution-connected resources will have played by distribution companies in the deploy- a stronger presence in energy markets, and their 3 ment of advanced metering, as well as data col- active contribution to the secure system operation lection and management. These questions are will be increasingly necessary through the provi- essential to unlock the flexibility potential of de- sion of ancillary services and network support. mand response, ensure a sustainable develop- In this new environment, distribution companies ment of self-consumption and facilitate well-func- can enable the participation of DER in compet- tioning retail markets. itive markets in a transparent and non-discrimi- Lastly, distribution companies may play a role in natory way. Moreover, the companies themselves the deployment and operation of new grid-edge are bound to become buyers of these flexibility infrastructures, such as public EV charging sta- services in order to ensure efficient distribution tions or distributed storage. The major regulato- planning and operation. Hence, it can be said that ry question is whether to consider this infrastruc- distribution companies will need to shift their role ture as part of the business model of distribution from that of network manager to market facilita- companies or, on the contrary, open them for pri- tor and system operator. vate initiative. It is noteworthy that, starting with similar goals The development of EVs and its impact on the (efficient DER integration, enhanced flexibility), distribution activity is yet one more signal that different regions may follow significantly differ- the integration of RES should be part of a holis- ent strategies. This is clearly illustrated by the fol- tic approach to achieving a low-carbon energy lowing two examples. system. The electricity system is more and more In the European context, regulators and poli- related to other sectors, such as transport and cy makers have placed a strong emphasis on heating/cooling. This requires policy makers to full market liberalisation. Therefore, distribution consider coherent and co-ordinated intersector companies or DSOs are prevented from per- planning. forming any activity potentially subject to com- 3.5.1 Moving from network managers petition. In those areas where this is not clearly to market facilitators defined – e.g., energy efficiency advice, distribut- and system operators ed storage ownership or metering data management – a strong regulatory supervision is advocated. Moreover, the enforcement of unbundling Regulation should grant DER access to rules is seen as a key policy measure to ensure a upstream energy and ancillary services, and transparent and non-discriminatory market func- facilitate greater coordination with TSOs/ISOs. tioning (CEER, 2015b). Meanwhile, unbundling Distribution companies should facilitate this rules make managing a significant amount of dis- participation and carry out activities such as persed resources a complex task, as discussed an ex-ante technical validation and an ex-post below. verification of the provision of the service. In a second example, the ongoing regulatory reform in the state of New York actually en- The growing presence of more flexible and di- courages utilities to seek what they have called verse distribution network users forces distribu- market-based earnings. These are new revenue tion companies to reconsider their convention- streams stemming from new activities, subject al roles as mere network owners and operators. to competition or not, beyond their conventional As the decentralisation of the power system ad- network-related duties. Box 3.11 describes a pro- vances, DER will become increasingly important posal submitted by one of the state’s utilities com- players in the overall functioning of power sys- prising four programs to yield market-based earn-

124 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

The distribution company as a market facilitator Box 3.11 Market-based earnings and new roles Despite the power sector transformation, distri- of utilities in New York bution companies will inevitably remain regulat- As part of the ongoing regulatory reform in the ed networks. Neutrality and transparency should state of New York, utilities are requested to submit govern any interaction between distribution com- to the regulator proposals for innovative business panies and network users. The tasks of regulators models, consistent with the sector transformation, and policy makers is to set the rules that encour- that can provide utilities with new revenue streams age network companies to act as neutral mar- – or so-called market-based earnings. For example, ket facilitators (CEER, 2015b). In this context, the Central Hudson Gas and Electric Corporation has term “market” comprises retail markets, energy presented a proposal including four of these pro- markets and ancillary services markets (THINK grams (PSC, 2014). Project, 2013). 1. A community solar program would install and op- Concerning retail markets, questions on the new erate a utility-scale solar plant and sell solar ener- role of distribution companies arise when the lib- gy, in blocks of 100 kilowatt hours (kWh) and at a eralisation process has been introduced in this fixed rate, to end consumers and energy services segment. On the one hand, distribution compa- companies willing to buy them. nies play a key role in managing the connections 2. A demand-response programme would require (in the case of new users) and disconnections (in end-consumers or aggregators on their behalf to case of contract termination or non-payments reduce their consumption upon request from the among the suppliers’ customers). This is a core utility. The goal is to reduce the state-wide peak function of distribution companies, which in a lib- demand and defer network investments in areas eralised context requires the intermediation of selected by the utility. suppliers who are the entities with a commercial 3. New micro-grids, targeted at consumers or relationship with end consumers. groups of consumers above 500 kilowatts (kW), The most critical role concerning a transparent would provide end-users with improved reliabili- and well-functioning retail market is that of man- ty in exchange for a given fee. aging metering data and providing data access 4. Voluntary smart meters would provide consum- to different stakeholders, particularly after the ers with enhanced information on consumption deployment of AMI. In a liberalised retail mar- patterns and pricing options to manage their en- ket, providing data access to retail companies in ergy bills. Subscribers would pay for the incre- a transparent and non-discriminatory manner is mental cost of the smart metering system. essential to ensure competitive market functioning. Moreover, this information must be offered to end consumers so that they can actively engage ings. Distribution utilities will also play a key role and make better-informed contracting decisions. in facilitating the participation of other players in Overseeing the behaviour of distribution compa- these services. According to this vision, utilities nies on this matter is particularly relevant where would become distributed system platform pro- these companies belong to a vertically integrated viders (New York DPS, 2014). It is acknowledged utility undertaking active in the retail market or that greater regulatory oversight will be neces- when distribution companies act as default sup- sary to ensure transparency and prevent unfair pliers. Section 3.5.2 discusses this specific topic practices among utilities (New York DPS, 2015). in more depth. Concerning energy and ancillary services markets, questions on the new role of distribution companies arise because of the participation of distribution-connected DER in these markets

125 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

that are run by a market or system operator (see Table 3.4). This participation is to be encouraged, Box 3.12 Balancing generation and demand with distributed flexibility 3 especially as the power system is more and more decentralised. Therefore, regulation should en- In Belgium, distributed energy resources (DER) are sure that the role of each stakeholder involved is already offering their flexibility to the transmission clearly defined, particularly that of the distribu- system operator (TSO) for balancing generation and tion company. demand. The distribution company performs a preliminary assessment of those users willing to provide More specifically, distribution companies may the balancing service in a prequalification stage and act as market facilitators by technically validat- send metering data to the TSO to verify the service ing the offers submitted by DER to the upstream provisions from DER (ISGAN, 2014). markets, i.e., by ensuring no distribution network The country plans on enhancing the transparency of constraints are violated. This role would be sim- this mechanism by setting up a real-time balancing ilar to what TSOs/ISOs do today with wholesale platform to allow the TSO to contract flexibility, sub- markets results. Additionally, after the service de- stituting the current bilateral agreements. Moreover, livery, distribution companies may be required to the distribution company is to be given more visibil- verify the provision of the services using meter- ity over the actions of DER and contracting options. ing data from DER. Box 3.12. presents the case of Lastly, if the flexibility provision of DER keeps grow- Belgium, where distribution-connected resourc- ing, its impact on the planned schedules of balances are already providing flexibility services to the ing responsible parties should be better considered. TSO. This case study also shows that the conven- Otherwise, the flexibility of DER would cause a de- tionally limited interaction between distribution viation in their control areas, for which they would companies and system operators needs to be re- be penalised. vised.

Interactions between the distribution company and the TSO/ISO areas of competence. Transmission and distribution networks interface in specific substations Under liberalised electricity markets, system op- that interconnect them. Under the current prac- erators (TSOs or ISOs) and distribution network tices, both operators interact periodically when operators have been unbundled. Both are in they plan the need for new network assets. With- charge of ensuring system secure operation and in the traditional, centralised, unidirectional flow adequate investment planning in their respective

Table 3.4 Major services that DER may provide to DSOs and TSOs

Type of DER able System operator procuring Service to offer the service such services

System balancing services All types of DER TSO Frequency control All types of DER TSO Voltage control All types of DER DSO Blackstart Larger-scale DS and DG TSO and DSO Short-term security DG, DS, DR, (EV) TSO and DSO congestion management

Source: THINK Project, 2013 Note: This table classifies DER into general categories, namely, DG, demand response, EVs and distributed storage. However, when the table states that a certain type of DER can offer a specific service, this does not mean that any member of that category is necessarily capable of doing so. For instance, the fact that DG, in general, could provide balancing services does not necessarily mean that any type of DG is technically capable or that it is economically reasonable for it to provide this service.

126 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

model, distribution companies provide TSOs with In this example, it is proposed that the DSO could forecasts of the load growth at their respective use the flexibility provided by customers with network interface points. DER (demand-side response, distributed generation, storage) to reduce the load condition of the With the change in paradigm driven by more TFO (Figure 3.15). There are other operating situ- flexible and decentralised resources connected ations – such as line congestion, voltage support directly to distribution networks and the more or black-start – where this co-ordination would active role of distribution companies operating be beneficial. those resources, there is an increasing need for co-ordinating actions between TSOs and distri- Those co-ordinated actions require DSOs to im- bution companies at the operational level. The plement innovative technology solutions that are flexibility connected at the distribution level may available but not yet deployed, such as grid mon- be an efficient resource for solving network prob- itoring, two-way communications with flexible lems, not only at the distribution level but also in customers and with the TSO, and network quasi the transmission network. real-time simulations.

Several models enabling this co-ordination be- The distribution company as a system operator tween the distribution and wholesale levels can The previous example of TSO-DSO interactions be envisioned. All these typically require some represents a case where the distribution compa- form of aggregation of a large number of DER, ny ought to request flexibility services from its either by the distribution company itself (which network users. Nonetheless, this would not be may be hampered by unbundling rules) or by the only case where the distribution company competitive agents such as retailers and aggre- should interact with DER to optimise the opera- gators who deliver services at both the distribution of the grid. There are situations in which this tion and wholesale level (IEA, 2016). In this re- would be desirable. Regulatory mechanisms that gard, ISGAN (2014) provides country examples of aim to foster this interaction include non-firm interaction between TSOs and distribution com- connection agreements, bilateral flexibility con- panies in grid operation are identified. Table 3.5, tracts and local markets (see Section 3.2.2 and for instance, shows the current practices and the Eurelectric, 2013; CEER, 2015b). future needs for TSO-DSO co-ordination in case of congestion in the transformer at the interface A relevant initiative in this regard can be found between the transmission and distribution net- in the European context. The new proposal for works when the transformer (TFO) is owned and an Electricity Directive issued by the European operated by the TSO. Commission on November 2016, if approved as is, would mandate Member States to ensure that reg-

Table 3.5 Current and future TSO-DSO interactions

Today Future

· Avoided in many countries by considering · More grid monitoring and intensified data n-1 criteria in the network planning exchange would allow using flexibility on the distribution grid to reduce · Cooperation mostly during transformer loading when necessary TFO congestion the planning phase · A request sent from the TSO · Emergency situations: TSO disconnects to the DSO could translate this request to distribution feeders, possibly through use-of-flexibility requests to flexible a request to the DSO customers connected to the distribution grid

Source: ISGAN, 2014

127 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Figure 3.15 Process for avoiding TFO congestion using flexibility In the distribution grid

3 Party Input Analysis Output

Required TSO TFO load monitoring TFO loading decrease

· Grid configuration Distribution of · Grid load monitoring Flexibility request requested flexibility, · Monitoring of flexible to single flexible DSO taking into account customers: actual customers network limits avaible flexibility

Source: Adapted from ISGAN, 2014

ulation enables and promotes distribution compa- treatment, promote competition when there are nies to procure flexibility services from network us- several potential suppliers for the same service ers (European Commission, 2016b). Moreover, this and reduce the potential for market power abuse Directive calls for the use of “transparent, non-dis- when the number of suppliers is low. criminatory and market based procedures” and requires distribution companies to define stan- 3.5.2 Advanced metering dardised and technology-neutral market products. to enable demand response and competitive retail markets This constitutes a deep change in the conven- Traditionally, small residential and commercial tional role of distribution companies, whose in- consumers have been offered very limited tariff teraction with end-users has been conventional- options with scarce time differentiation. However, ly limited to grid connection and disconnection, the integration of large shares of RES, increased management of supply outages and, where the interest from consumers in making better-in- distribution company was part of a vertically informed decisions and the efforts to introduce tegrated utility, billing. Thus, distribution compa- competition in the retail electricity sector re- nies ought to evolve from purely network manag- quire a shift in paradigm. Advanced metering is ers to true distribution system operators, actively a key enabling technology for the active partic- managing the resources connected to their grids, ipation of electricity end-users. Thus, smart me- similar to what TSOs do today. Moreover, new tering is a prerequisite for demand response and market players, such as aggregators, may come well-functioning retail markets, as well as a sus- on the scene, who combine the responses from tainable penetration of self-consumption and an individual customers responding to the needs of active management of EV charging and distribut- the distribution company. ed storage. In the new context concerning liberalised mar- There are two main policy and regulatory deci- kets, regulation should enable the change and sions in this domain: 1) ownership and manage- clearly define the roles and responsibilities of dis- ment model for the deployment of advanced me- tribution companies and market players. Regula- ters and 2) management of metering data and tory oversight will be needed, particularly where granting access to stakeholders in a transparent the distribution company belongs to a larger ver- and non-discriminatory way. tically integrated company, to avoid preferential

128 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Advanced metering roll-out tive cost-benefit analysis is obtained or where The installation and management of metering a large-scale rollout is not planned. This meter equipment have been conventionally carried out must be provided under fair conditions and com- by distribution utilities, especially for small and ply with minimum technical standards defined in medium consumers. Thus, metering assets have the proposal for a new Electricity Directive (Euro- been traditionally treated as part of the monopo- pean Commission, 2016b). Table 3.6 summarises listic activities of distribution companies who re- the cost-benefit analysis results obtained for Ger- cover the corresponding costs through network many and the United Kingdom. charges or a rental fee. A model where the dis- According to a benchmarking study carried out tribution company is responsible for the deploy- by the European Commission, as of mid-2014, ment of advanced metering seems the most im- 16 countries had decided to implement a large- mediate solution, even in contexts where retail scale deployment, another three had opted for and distribution have been unbundled. For ex- selective deployment, and just four had rejected ample, in the European Union, distribution com- advanced metering altogether (European Com- panies are responsible for meter installation and mission, 2014a). A potential drawback of a large- ownership in most member states, with just a few scale rollout is that some consumers may op- exceptions, such as in Germany and the United pose it due to alleged health or privacy concerns. Kingdom (European Commission, 2014a; 2014b). Therefore, some regulators have decided to in- In Germany, consumers are entitled to choose troduce an opt-out clause in the advanced meter- any metering operator, i.e., metering is legally ing programs, as in the case of California (CPUC, considered a fully competitive activity. Nonethe- 2012a; 2012b) and the The Kingdom of Nether- less, distribution companies are still the default lands (European Commission, 2014c). metering operators and deliver metering services Brazil offers a middle path between a mandat- to those consumers who do not explicitly opt for ed rollout and complete free customer choice. a third-party metering operator. Meanwhile, elec- In 2012, the regulator passed a norm mandating tricity metering in the United Kingdom falls under distribution companies to offer their customers the responsibility of suppliers (European Com- the choice of installing a smart meter. The offer mission, 2014a). highlighted the potential benefits of access to en- In order to foster the penetration of advanced hanced information, more tariff options and re- metering and benefit from economies of scale, mote connection management (ANEEL, 2012). policy makers may decide to opt for a large- Advanced metering data management scale rollout. This process usually stems from a and access policy mandate on metering operators, normal- The adoption of advanced metering represents a ly including rules governing this deployment in revolution in terms of the volume of information it terms of schedule or technology capabilities. For makes available on consumer behaviour and net- instance, EU Directive 2009/72/EC19 mandates work utilisation. Further, smart meters may re- European countries to install smart meters in at cord not only consumption data, but also techni- least 80% of consumers by 2020, provided a pos- cal information on power quality levels, outages, itive benefit-cost analysis is obtained (European meter tampering, etc. Hence, advanced metering Commission, 2009). Acknowledging the key role data have immediate application to distribution that smart metering plays in terms of consum- network operations and planning, and distribu- er awareness and retail market functioning, the tion companies should be granted access. European Commission has proposed to provide consumers with the right to have a smart meter The main policy and regulatory concerns around installed even in those countries where a nega- advanced metering lie on the commercial value

19. Directive 2009/72/EC of the European Parliament and of the Council of 13 July 2009 concerning common rules for the internal market in electricity and repealing Directive 2003/54/EC (Text with EEA relevance).

129 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Table 3.6 Smart metering cost-benefit analysis (CBA) in Germany (left) and the United Kingdom (right) 3 POSITIVE CBA OUTCOME (for the Roll-out Scenario Plus) CBA OUTCOME POSITIVE NEGATIVE for the EU scenario

· € mn 6,493 (by 2022) Total Investment Total Investment € mn 9,295 · € mn 14,466 (by 2032)

· € mn 5,865 (by 2022) Total Benefit Total Benefit € mn 21,749 · € mn 16,968 (by 2032) Cost/metering point (as communicated €546 Cost metering point €161 by the Member State)

Benefit/metering point Benefit €493 €377 (as communicated by per metering point the Member State)

Consumers’ benefit Consumers’ benefit 28% (domestic sector) and 60% 47% (% of total benefits) (% of total benefits) (non-domestic sector) Domestic sector (electricity + gas) · Supplier cost savings (54%) · Energy savings – 33% · Energy savings (28%) Main benefits · Load shifting – 15% Main benefits · Carbon savings (7%) (% of total benefits) (% of total benefits) · Avoided investments in the distribu- Non-domestic sector (electricity + gas) tion grid – 13% · Energy savings (60%) · Carbon savings (19%) · Supplier cost savings (15%)

Domestic sector (electricity + gas) · Smart meters CAPEX+OPEX (43%) · Investments smart metering systems · Communication costs CAPEX+OPEX (meter, gateway, communication in- (23%) Main costs frastructure) – 30% Main costs · Installation costs (15%) (% of total costs) (% of total costs) · Communication costs – 20% Non-domestic sector (electricity + gas) · Smart meters CAPEX+OPEX (49%) · IT-costs – 8% · Communication costs CAPEX+OPEX (31%) · Installation costs (16%)

Energy savings Energy savings (% of total electricity 1.2% (% of total electricity 2.2%; gas 1.8% consumption) consumption)

· 0.5% - 1% (as a percentage · 1.3% in average between 2014 and Peak load shifting Peak load shifting of total consumption) (% of total electricity 2022 (% of total electricity · 1.3% - 2.9% (as a percentage consumption) · 2.9 in 2032 consumption) of peak consumption)

Source: European Commission, 2014c Note: The “EU Scenario” analysed in Germany corresponds to the requirement set in Directive 2009/72/EC, i.e., to install smart meters for at least 80% of all consumers by 2020. The Roll-out Scenario Plus considers the installation of smart metering systems in new and existing RES and CHP units with a contracted power above 250 watts (W), whereas it limits the installation of smart metering by 2022 to large consumers and new or renovated buildings (remaining consumers will be equipped with meters without external communications).

130 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

of this data, with respect to retail market func- third party driven by the size and structure of the tioning and the provision of energy services. distribution sector? Regardless of the model cho- For companies contracting with end-consum- sen, regulators ought to ensure that data access ers (suppliers, energy services companies, ag- is granted in non-discriminatory conditions, while gregators), this information has commercial val- the privacy of end-users is ensured. ue. Consumers, meanwhile, may be able to make Another issue that has attracted policy attention better-informed decisions (CEER, 2015a). Princi- is the format in which these data are stored and ples guiding the organisation and regulation of shared. To remove market barriers and reduce data access seek to ensure consumers’ privacy administrative costs, Article 24 of the proposal and data transparency, accuracy, accessibility as for a new Electricity Directive in Europe (Europe- well as the avoidance of discrimination (CEER, an Commission, 2016a) states that Member States 2015a). have to define a common data format. Moreover, These topics are relevant in those countries where the European Commission could define a manda- the retail market has been liberalised, especially if tory Europe-wide data format as well as data ac- the regulator is concerned about insufficient un- cess procedures replacing national approaches. bundling or vertical integration. Note that even in those countries where a distribution company is 3.5.3 Distributed storage responsible for meter deployment and operation, and ownership models providing access to these data and managing that The reduced cost of energy storage battery sys- access is not necessarily the task of distribution tems, combined with the need for enhanced flex- companies (CEER, 2015b). Thus, despite the fact ibility in the distribution network, opens the pos- that distribution companies have conventionally sibility of deploying small- and medium-sized been responsible for metering activities, they will distributed storage for grid support. Interest in not necessarily act as AMI data managers: such applications is reflected in the large number of demonstration projects being implement- “CEER remains of the view that there is a need ed worldwide.20 Battery systems may be installed for a neutral data coordinator or data hub to both on the premises of end-users or directly manage and provide access to data, and that connected to the distribution grid. this role can be provided by a number of different parties as is already the case in some EU The key regulatory questions here include: 1) countries. […] CEER believes that DSOs should whether to allow the distribution utility to own remain as neutral market facilitators but that and operate the storage system, given that this this does not automatically confer the status may collide with existing unbundling rules and 2) of data management coordinator to a DSO” how to ensure that the storage is located where (CEER, 2015b: 13). it is most beneficial to the distribution network when the distribution company cannot own and The European Smart Grid Task Force identified operate the storage system. three main models for the management of smart metering data: centralised management operat- Energy storage systems can provide grid support ed by the distribution company, centralised man- services to distribution companies as well as oth- agement operated by an independent regulated er system services, such as balancing or price ar- agent and a decentralised model using a data bitrage (Eurelectric, 2012; THINK Project, 2012; US access-point manager (Smart Grids Task Force, DOE, 2013). Box 3.13 offers a benefit-cost analysis 2013). All of these models have pros and cons. of a grid-connected battery system, illustrating They also involve many questions. For example, the challenges faced by this type of application. is the decision on whether to allocate the role of The benefits from distribution grid support alone data manager to the distribution companies or a may not be enough to result in a positive business

20. A comprehensive database of storage projects can be found at: http://www.energystorageexchange.org/projects/.

131 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box 3.13 The benefit-cost analysis of a grid-connected storage system

The Public Service Company of New Mexico (PNM) has technically. However, as shown in Table 3.7, the bene- 3 carried out a demonstration project with a battery energy fit-cost analysis yielded a negative result, even including storage system used to smooth the impact on a photovol- benefits that would not normally go to the distribution taic (PV) power plant through peak shaving and voltage utility, such as deferred generation capacity or emissions control in a medium-voltage (MV) utility feeder. The sys- reduction. The same report states that the benefits to the tem schematic is shown in Figure 3.16 distribution operator could have been much higher if the storage system had been located in a feeder with a higher An evaluation of system performance in the period Sep- PV penetration. tember 2011 until February 2014 is presented in PNM (2014). The results show that the system performed well

Figure 3.16. One-line diagram of PNM’s battery storage pilot project

PNM Distribution

835 kVA, 750 kW BES System supplied by BESS Master Grid-Tied Inverter EPM/Ecoult 500 kW Central Inverter Fixed Voltage 800 VDC (+/- 400 VDC)

500 kW 250 kW DC Converter with DC Converter with Power Regulator Power Regulator

500 kWp Solar PV Array Power Smoothing - 500 kW Peak Shifting - 990 kWhr Containerised VRLA Containerized Advanced Ultra Batteries (2xCABS) Carbon VRLA Battery (6xCABS)

Source: Adapted from PNM, 2014

Table 3.7 Summarised benefit-cost analysis of PNM’s battery storage pilot project

Costs Benefits

Utility Revenue Requirement (Variable) $51,576.11 - Utility Revenue Requirement (Fixed) $2,929,123.43 - Electricity Sales - $114,735.61 Distribution Investment Deferral - $333,987.30 Base Case Distribution Losses Reduction - $15.80 System Electric Supply Capacity - $177,036.26 Emmission Offset $45,583.56 Total $2,980,699.54 $671,358.53

Source: PNM, 2014

132 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

case, and the benefits to the grid largely depend storage assets beyond the demonstration proj- on the location of the storage system. ects (AEEGSI, 2015a). As shown in Figure 3.17, these conditions would be limited to non-com- Existing applications for grid support, usual- petitive activities or to small-scale applications, ly within demonstration projects, mainly follow and in all cases subject to a benefit-cost evalu- a structure where the distribution company di- ation following a methodology approved by the rectly operates the storage system. However, regulator. this may not be a viable model in regions where regulation mandates the legal unbundling of Meanwhile, the European Commission has stat- power distribution. Utilising the storage system ed its will to forbid distribution companies from strictly for grid support would lead to largely owning, developing, operating or managing stor- underutilised systems, given that network con- age facilities (European Commission, 2016b). Ex- straints in a specific area may arise only a few emptions to this rule may be implemented if the hours per year. On the other hand, the provision following three criteria are met: 1) other parties of system services subject to competition would do not express interest in these activities in an result in distribution companies acting de facto open tendering procedure, 2) storage facilities as market agents. enable distribution companies to fulfil their obligations and 3) the regulator verifies compliance In order to address this problem, the Italian reg- with the previous two requirements and pro- ulator has recently proposed a framework to de- vides its approval. In order to ensure that regula- termine the conditions under which distribution tion keeps up to date with technology and busi- operators may be allowed to own and operate

Figure 3.17 Framework proposed by the Italian regulator to decide upon the role of distribution companies in storage ownership

Are the rules enabling DG to take part to the ancillary service market deﬁned?

no yes

Is the storage application connected to MV network? Is the storage application below no (LV) the minimum power threshold?

A simpliﬁed CBA methodology is envisaged for LV applications yes yes

Is the DSO able to demonstrate, through a CBA case (ex-ante approved methodology), no the cost-eectiveness of this storage application?

yes no

ALLOWED NOT ALLOWED

Source: Adapted from Lo Schiavo, 2015

133 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

ness model developments, the Commission has need to justify their contribution to the following also proposed that regulators who allow distri- goals: grid optimisation, integration of renew- 3 bution companies to own storage facilities pe- able energy or a reduction of greenhouse-gas riodically (every five years) re-assess the pre- emissions. The goal of this policy decision is to vious conditions through a public consultation spur the adoption of energy storage. Therefore, and modify regulation accordingly if needed. despite the fact that utilities are given a central Another interesting initiative to encompass var- role in the elaboration or procurement plans, they ious phases from storage demonstration to de- may retain the ownership of no more than 50% of ployment is seen in California, where authorities the storage capacity. have set binding targets on the three largest in- An auction-based mechanism similar to the one vestor-owned utilities to deploy 1 325 MW of stor- being implemented in California could also be ap- age capacity by 2020 (CPUC, 2013). The target is plied in countries like Italy, where unbundling is broken down per utility and type of application, in place, as a means to ensure storage systems distinguishing among grid-connected systems, are located where they are most needed by the either transmission or distribution, and customer distribution company. The distribution compa- storage (see Table 3.8). ny may determine the location and grid-support In order to comply with this requirement, utilities service required, and the winning storage oper- shall carry out competitive solicitations every two ators would engage in long-term contracts with years. Before these tendering processes, utilities the distribution company that would contribute are required to submit procurement plans to the to their business cases. regulatory commission for their evaluation, which

Table 3.8 Proposed storage targets for IOUs in California (MW installed per year)

Storage Grid Domain Point 2014 2016 2018 2020 Total of Interconnection

Southern California Edison

Transmission 50 65 85 110 310 Distribution 30 40 50 65 185 Customer 10 15 25 35 85

Subtotal SCE 90 120 160 210 580

Pacific Gas and Electric

Transmission 50 65 85 110 310 Distribution 30 40 50 65 185 Customer 10 15 25 35 85

Subtotal PG&E 90 120 160 210 580

San Diego Gas & Electric

Transmission 10 15 22 33 80 Distribution 7 10 15 23 55 Customer 3 5 8 14 30

Subtotal PG&E 20 30 45 70 165

Total – all 3 utilities 200 270 235 490 1,325

Source: CPUC, 2013

134 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

3.5.4 Business models for the To conclude, it is useful to recall the two main deployment of electric vehicle regulatory questions posed at the beginning of charging infrastructure this section: 1) whether to allow distribution companies to own and operate storage assets and 2) how to ensure an appropriate siting and sizing of As market forces alone may not be enough storage units from the viewpoint of the distribu- to foster public charging infrastructure at early tion grid. As discussed throughout this section, stages, policy makers and regulators may have enabling distribution companies to install and to step in to kick-start the EV sector. operate storage assets would facilitate the use of storage for network support since distributors Electric mobility has been identified as a key fac- would directly decide its location and operation. tor in the reduction of carbon emissions and local Nonetheless, this may lead to under-utilised stor- pollution. Hybrid vehicles, which utilise the elec- age assets since network constraints usually oc- tricity produced on-board, have been in commer- cur only on occasions. Moreover, grid-support cial operation for quite some time. Nonetheless, services are not normally enough to yield a posi- the full benefits of e-mobility require the devel- tive business case for storage devices. Therefore, opment of plug-in EVs, which recharge their bat- storage operators would probably need to seek teries by connecting to the power grid. additional revenue streams, such as price arbitrage or balancing services, which are sometimes Charging points may be located in a wide variety forbidden for distribution utilities due to unbun- of places such as at homes, working places, park- dling or market access rules. ing lots or shopping centres. The deployment of a network of public charging points would en- On the other hand, leaving the decisions on stor- able the large-scale adoption of EVs. Distribu- age siting and sizing to market players would not tion companies do not play any role concerning ensure that this is located in the place and in the charging points in private areas besides granting amount needed from the network perspective. access to the distribution grid on the same con- The distribution company would need regulato- ditions as any other network user. Nevertheless, ry mechanisms to overcome these barriers. Thus, they may play a more active role in the deploy- regulators may implement exemptions on the un- ment of public charging infrastructure.22 bundling obligations under specific circumstances or enable distribution companies to contract Public charging services can be considered a services with storage operators through tender- competitive activity subject to pricing alterna- ing schemes. Concerning the operational deci- tives such as prepayment, flat monthly rates, free sions over storage systems for the provision of charging to promote another product or service, network support (and preventing storage sys- etc. For example, in the United States, several dif- tems from causing grid constraints), distribution ferent charging suppliers provide their services companies may purchase such services through throughout the country. According to the alter- the mechanisms discussed in Sections 3.2.2 and native fueling station locator23 from the US DOE, 3.5.1, which enable them to become true system there are more than 11 000 EV public charging operators.21 stations across the United States, of which less than 1.5% are directly operated by the utility. The

21. Note that, as discussed in Section 3.3.2, distribution companies should be encouraged by regulation to reduce overall costs, without favouring OPEX over CAPEX reductions. Otherwise, they would see little incentive to rely on storage as an alternative to network investments or, provided they are allowed to own storage, they may overinvest in storage assets to increase their regulatory asset base, benefitting from information asymmetries. 22. The deployment of (plug-in) EVs also affects the distribution grids planning and operation likewise DG, as discussed in Section 3.2. Moreover, distribution companies may resort to the flexibility potential offered by EV charging points, either when these are decentralised (e.g., in-home charging posts) or centralised (charging stations). Thus, a large-scale adoption of electromobility will stress the need for distribution companies to adopt their new role as SOs (Section 3.5.1). 23. Accessed in October 2015: http://www.afdc.energy.gov/locator/stations/.

135 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

remaining stations are operated by private net- by unbundling rules. For instance, the Europe- work charging suppliers or public authorities, as an Directive 2014/94/EU24 on Alternative Fuels 3 described in Box 3.14. Infrastructure states that the operation of public charging points is a liberalised activity and Market forces alone may not succeed in deploy- distribution operators ought to co-operate on a ing this infrastructure where market size is lim- non-discriminatory basis with any charging point ited (THINK Project, 2013). Policy makers may operators (See Box 3.15). Moreover, a deployment need to kick-start the EV sector in its initial stag- led by the distribution company would imply that es, for example, by allocating this role to distribu- rate payers would be subsidising EV users. tion companies and treating EV charging points as regulated assets. But this may be hampered

Box 3.14 Business models and drivers for electric vehicle public charging in the United States

Tesla Motors, a leading electric vehicle (EV) manufactur- overcome range anxiety concerns and promote EV sales. er, has deployed one of the most widespread networks The remaining major network charging suppliers (Sema- of public fast charging points based on its so-called su- Connect, ChargePoint or Blink)a are subscription-based percharger. Tesla superchargers across North America companies that purchase electricity from power utilities are shown in Figure 3.18. and sell it to their subscribers. These networks also sell Tesla Motors follows a strategy based on placing their charging points to private customers, having a range of superchargers along transited highways and in congested different products for domestic users, commercial cus- city centres, where users may charge their vehicles for free. tomers and others. Thus, their business model is based The company’s motivation, as a vehicle manufacturer, is to both on the sales of electricity and charging points.

Figure 3.18 NetworkSeward of Tesla Motor’s superchargers across North America

Juneau

Prince Edmonton George

Calgary Regina Winnipeg Vancouver

Victoria Thunder Bay Seattle Bismarck Duluth Québec Helena Charlottetown Fargo Fredericton Portland Billings Montréal Minneapolis Salem Pierre St-Paul Montpelier Augusta Halifax Boise Toronto Milwaukee Portland Albany Bualo Boston Detroit Omaha Cheyenne Cleveland New York Des Moines Chicago Providence Lincoln Pittsburgh Salt Lake City Denver Kansas City Indianapolis Philadelphia Carson City Colombus Topeka Cincinnati Baltimore Sacramento Jeerson Wichita City St-Louis San Francisco Richmond Las Vegas Norfolk Nashville Oklahoma Memphis Santa Fe Charlotte Raleigh City Little Rock Albuquerque Chattanooga Los Angeles Columbia Phoenix Birmingham Atlanta Fort Worth Dallas San Diego Jackson Charleston Mexicali El Paso Montgomery Ciudad Juárez Baton Rouge Austin Tallahassee Mobile Hermosillo Houston San Antonio New Orleans Open Now Tampa Monterrey Opening soon Miami Culiacán

Tampico Mérida Santiago Guadalajara Cancún Campeche de Cuba Morelia SanJuan Acapulco

Source: Adapted from Tesla motors, 2017 Cartagena Maracaibo Barquisimeto

Ciudad Guayana a. http://www.semaconnect.com/; http://www.chargepoint.com/; http://www.blinknetwork.com/. Medellín

24. Directive 2014/94/EU of the European Parliament and of the Council of 22 October 2014 on the deployment of alternative fuels infrastructure Text with EEA relevance.

136 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

Box 3.15 Charging service model according to the European Directive 2014/94/EU on Alternative Fuels Infrastructurea The development of electric mobility requires a careful −− The electricity distributor must operate in a nondiscrimi- regulation of the contractual relationship between the natory way, both in providing the connection service [4] to various subjects involved: electricity distribution opera- the CPO and the electricity transport service to the elec- tors, electricity suppliers, charging point operators, mo- tricity supplier chosen by the CPO [3]; in order to ensure bility service providers, mobility service providers (for the electricity distributor’s impartiality and efficiency, ap- instance, EV manufacturers or other specialized subjects) propriate unbundling rules should apply where the distrib- and EV drivers. utor belongs to a vertical integrated group that also provides (with undertakings separate from the distribution As shown in Figure 3.19, the European Directive 2014/94/ operator) the electricity supply and/or charging activity. EU on Alternative Fuels Infrastructure (AFID) sets some principles that address these contractual relationships −− The CPO shall be free to choose one or more power sup- with the aim of creating an open competitive market for pliers from any supplier in the Union [2] and must provide recharging services. The basic assumption is that the a charging service to EV drivers [1], providing that pay- good supplied is the bundled “recharge” product; the ment of the service can be, as chosen by EV drivers, ei- EV driver should not, therefore, separately purchase the ther directly [1a], with the same modes of fuel supply, or charging service and the electricity needed. The final cus- via the intermediation [1b] of mobility service providers, tomer of the power system is the charging point operator typically EV manufacturers or other specialised entities. (CPO), who has a supply contract with an electricity sup- Further, the EV driver must be able to recharge even plier, freely chosen in the retail market. The EV driver has −− without having a contractual relationship with the CPO, no contractual relationship with the electricity supplier. on ad-hoc basis (single transaction settled with ordinary Some of the key guidelines provided by AFID are: payment means).

Figure 3.19 Contractual relationship between actors involved in electric mobility

MOBILITY SERVICE 2 ELECTRICITY 1b PROVIDERS (MSP) SUPPLIER CHARGING POINT OPERATOR (CPO) 1b 1a EV DRIVERS Charging point 3 4

Charging Electric point vehicles (EVs) ELECTRICITY DISTRIBUTION Charging COMPANY point Point of Displays delivery (can be set Distribution grid Other uses to zero at each of electricity transaction) Smart meter Charging station

Legend: 1a: EV Driver –Charging Point Operator (directly); 1b: EV Driver – Charging Point Operator (intermediated by Mobility Service Provider); 2: Charging Point Operator – Electricity Supplier; 3: Electricity supplier – Electricity distribution company; 4: Electricity distribution company – Charging Point Operator (only for connection) a: Directive 2014/94/EU of the European Parliament and of the Council of 22 October 2014 on the deployment of alternative fuels infrastructure. Source: Adapted from Lo Schiavo, 2017

137 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Other ways to provide the policy push could be Distribution companies ought to facilitate the explored. For instance, local governments or oth- participation of DER in energy and ancillary ser- 3 er public entities may allocate the deployment vice markets. In this regard, distribution compa- and operation of charging points through a long- nies may validate the technical feasibility of the term contract assigned in a tendering process offers submitted by DER to the upstream mar- (THINK Project, 2013). Thus, the role of the au- kets and verify the provision of the services ex thorities would consist of setting the conditions post through metering data. In order to carry out of such a contract (number of charging points, these duties, the conventionally limited interac- pricing rules, standardisation requirements, etc.) tion between distribution companies and system and organising the auction. operators needs to be revised. Co-ordinating the actions of TSOs and distribution companies at the 3.5.5 Conclusions and operational level requires the implementation of recommendations innovative technology solutions. As the energy transition evolves, a growing share In addition to market facilitators, distribution of the generation capacity and the flexibility re- companies themselves should make use of DER sources needed to ensure the secure system op- flexibilities to ensure efficient distribution plan- eration will be connected to the distribution net- ning and operations. Thus, enhanced interaction works. An increase in demand awareness and the with end-users is necessary so that distribution deployment of advanced metering infrastructure companies become true distribution system oper- is contributing to the change in the landscape ators actively managing the resources connected seen by distribution companies. In response, dis- to their grids. In this context, regulations should tribution companies in liberalised markets need define clearly the responsibilities of distribution to adapt their role to ensure the efficient integra- companies and oversee potential anti-competition of DER. tive behaviour, especially where the distribution In countries where policy makers pursue the liber- company belongs to a vertically integrated com- alisation of the electricity sector, distributioncom- pany in a context of retail competition. panies help create well-functioning retail mar- Lastly, regulation ought to clarify the role of dis- kets. A critical task related to the increase of DER tribution companies in the deployment and op- is the need to provide market agents with access eration of innovative infrastructures such as ad- to metering data in transparent and non-discrim- vanced metering, distributed storage or EV public inatory conditions. This might be seen as a con- charging infrastructure: ventional task of distribution companies. Howev- er, when these companies belong to a vertically •• Advanced metering is a key enabling tech- integrated undertaking active in the retail market nology for demand response, as well as for a or act as default suppliers, regulators may wish sustainable penetration of self-consumption to explore alternative data management models. and an active management of EV charging and distributed storage. A large-scale roll-out re- Different models can be found, depending on the quires a policy mandate setting the technical degree of the decentralisation of data manage- and economic conditions. Advanced metering ment and the involvement of distribution compa- deployment is normally performed by distribu- nies. There is no consensus on the most appro- tion companies, except in cases where compe- priate model. The final choice is mainly driven by tition has been introduced in metering services. the size and structure of the distribution sector Optionality clauses may be implemented as a in each country. In any case, regulators must en- response to opposition from end users due to sure non-discriminatory data access and protect privacy or health concerns. consumers’ privacy, particularly after the deploy- •• Distributed storage has the potential to supply ment of advanced metering. grid-support services. However, this does not automatically mean that distribution compa-

138 DISTRIBUTION NETWORKS AND DISTRIBUTED ENERGY RESOURCES

nies should own and operate storage devices. EV market development. Policy makers may In fact, unbundling provisions could rule this kick-start the EV sector by endowing distribu- possibility out due to the fact that storage tion companies with responsibility for its de- operators may need to provide other services ployment. However, in some context this may under competition in order to obtain a positive not be possible. On the one hand, unbundling business case. Thus, exemptions on the unbun- rules may prevent distribution companies from dling obligations may be necessary. Alterna- selling electricity to EV users. On the other tively, distribution companies may be entitled hand, treating EV charging points as part of the to contract services with storage operators regulated asset base would imply that rate pay- through tendering schemes. ers would be subsidising EV users. Thus, other •• Market forces alone may not foster public policy alternatives could be explored to provide charging infrastructure due to the low level of the initial policy push.

139 LESSONS LEARNT AND POLICY RECOMMENDATIONS CONCLUSIONS AND RECOMMENDATIONS 4

The world is in the midst of an energy transition. 4.1. WHOLESALE MARKET DESIGN Power sectors are experiencing – or will soon ex- Chapter 2 focused on the upstream segments of perience – profound changes that include low car- the power sector, more specifically on generation bon technologies in the generation mix; the de- and system operations, and in particular on the centralisation of generation capacity; an increase wholesale market design refinements needed to in regional inter-connection and market integra- lead the system expansion towards a more effi- tion; consumer empowerment, thanks to better cient and sustainable future. information on pricing alternatives and consumption profiles; and an increase in the availability The discussion is divided into two major sections: and utilisation of a diversity of distributed energy The first reviews the many important design ele- resources (DER). ments that need to be fine-tuned to improve the In response to the challenges posed by this trans- performance of short-term market mechanisms, formation, renewable energy policy and pow- from the day-ahead market to the balancing mar- er sector regulation must work in co-ordination, ket. The independent system operators of the and be adapted to the new reality, as discussed United States and the Power Exchanges of the in Chapters 2 and 3. European Union were used as case studies. The lessons extracted from these two can be, trans- To be effective, both regulation and policy must posed to other liberalised markets. be designed and implemented whilst taking into account the specific characteristics of each pow- The second focuses on long-term mechanisms er sector, such as the degree of its liberalisation designed by electricity regulators to intervene and its institutional arrangements, market struc- in markets with the objective of enhancing the ture, grid development, electrification rate and penetration of additional sources of capacity (in- renewables penetration. cluding RES), i.e., capacity remuneration and RES support mechanisms. The discussion follows two This section summarizes the main findings of this main perspectives. First, capacity remuneration volume and provides an overview of the major mechanisms are reviewed from the point of view policy and regulatory recommendations offered. of RES. Second, the main design elements of RES support mechanisms are evaluated, seeking the

141 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

best regulatory practices that minimise these In the European Union, day-ahead markets run mechanisms’ interference in the functioning of by power exchanges were originally envisioned 4 the markets, while at the same time maximise as simple electricity auctions. Using simple bids their efficiency as a tool to promote the installa- involves the need for market agents to anticipate tion of low-carbon technologies. the resulting dispatch and internalise, in the bids, all operational costs and constraints – a task that, 4.1.1. Adapting short-term market design to be properly fulfilled, requires that market con- Short-term energy market design needs to be en- ditions be predictable. hanced and refined at all levels, particularly with The growing penetration of VRE has increased respect to time frames, bidding formats, clearing uncertainty and this has called for greater com- and pricing rules and integration with reserves plexity in bidding formats. But instead of opting and regulation markets. for the multipart formats used in the traditional The type of refinements needed depends on the pool or in the United States ISOs, in the European specific market design approach. Differences in Union each power exchange has progressively in- the functions of the market operator (and within corporated the so-called block orders and semi- it, the Power Exchange) and the system operator complex orders. are key. These differences are what set apart the Multipart offers, while more complex than blocks United States and European Union market de- or semi-complex orders, are a more “natural” signs: the degree of separation between the roles way of representing the generation resources, of market operators and system operators, which and have advantages if a multi-part bid is able is not present in the United States, and quite to replace multiple blocks. However, including marked in the European Union. multipart bids in the European Union power ex- Bidding formats changes would require significant changes in the −− market design and pricing and clearing rules. In liberalised power systems in the United States, independent system operators (ISOs) require Continuous improvements in bidding formats are generators to submit multipart offers that try to expected in both market models, with the aim of faithfully represent the detailed operational (and achieving a more efficient integration of intermit- opportunity) costs as well as the technical content renewables. straints of their generating units. This bid format is robust against high penetration levels of re- −− Pricing and market clearing newables, because it allows generation agents to There are two ideal properties of pay-as-cleared be efficiently scheduled and dispatched. (uniform) marginal pricing: first, marginal pricing is consistent with optimal dispatch and, second, The capability of the ISO model to integrate these settling the transactions of all agents at the same new resources, however, could be further im- price (uniform pricing) sends an efficient signal proved. Resources with different characteristics for bidding at true costs and minimising ineffi- could be better integrated into the market. This ciencies. calls for a larger number of bidding formats tailored to specific needs. For example, regarding Perfect marginal pricing can be achieved only un- the relevant case of demand response, Liu et.al. der ideal conditions – in particular, the absence (2015) point out that “the bidding system does of non-linearities or “lumpiness” in the produc- not always provide a mechanism as an alternative tion cost functions, which are common in actual to the price-quantity bid format for consumers to markets. The presence of VRE generation exac- express their willingness to adjust consumption, erbates these complex behaviors. This requires particularly in response to price signals”. choosing between the objectives of achieving an optimal dispatch or implementing a uniform marginal pricing scheme in the clearing and pricing

142 CONCLUSIONS AND RECOMMENDATIONS

market rules. The United States has opted for the •• As mentioned in regard to bidding formats, first objective, while the European Union has cho- market clearing rules in both market models sen the second. seek to efficiently integrate larger shares of variable renewables. Their development goes In the United States, discriminatory side pay- hand in hand: bidding formats condition mar- ments, known as uplifts, are unavoidable ele- ket-clearing and pricing rules, and these rules ments of a pricing system consistent with a strict- in turn affect the bid formats. ly welfare-maximising dispatch. The underlying problem is that the uplifts are outside the uni- −− Locational granularity of prices form marginal price (or an ensemble of locational and schedules: zonal versus nodal marginal prices), and this hampers the develop- The increased deployment of VRE, particular- ment of a correct market signal that applies to ly wind, may result in a more constrained trans- all agents. The approaches to the uplift problem mission network. Zonal pricing is a simplified ap- reviewed in this report share something in com- proach to reflect network constraints, but it may mon: an attempt (never completely fulfilled) to come at the expense of market efficiency. In this internalise, as much as possible, all costs in uni- context, nodal pricing can often provide better form prices, i.e., to minimise the uplift to get as operation and investment signals, but it may be close as possible to uniform pricing. more challenging to implement. In the European Union, giving priority to the single-pricing rule leads to a short-term dispatch −− Rethinking reserve requirements that deviates from the cost-minimising one. As and procurement argued before, this is a matter of trade-offs; in System operators need to implement new solu- the European context, supporting uniform pric- tions to improve the reserves-supply function so ing is an objective worth the loss in short-term that they are priced according to the value they cost efficiency. There are three major problems provide to the system. In addition, energy and re- with the European Union day-ahead pricing and serve markets should be properly connected to clearing design: 1) relative lack of transparency in allow the former to reflect the actual conditions the algorithm,2) some units are unable to set the of the latter. market price and 3) the algorithm’s complexity In the reserve markets, quantity requirements limits the number of complex and block orders are typically procured through market mecha- that can be handled. nisms. The willingness to pay for each megawatt These three problems are all connected to the (MW) of reserve below a minimum contingency underlying complexity of clearing the market level is set to the value of lost load – although with uniform prices. Current efforts concentrate sometimes the price is artificially set below this. on improving the computational performance of Above this minimum contingency level, the mar- EUPHEMIA (the algorithm used, see Appendix) to ginal value of any additional MW of reserves is cope with the increasing number and complexity typically set to zero – a practice that should be of bids. questioned on the basis of rigorous regulatory principles (Hogan 2014). •• PCR-ESC (2015) proposes long-term solutions to solve the complexity issue, such as: Time frame of markets, dispatches and prices •• Reduce the amount of blocks types and other −− As variability increases, the time granulari- complex products allowed per participant and ty of market signals needs to increased. Better- market (bidding zones). adapted and more flexible market timelines are •• Reduce the range of products treated in EU- needed. In the European Union case, guarantee- PHEMIA. ing full liquidity in intraday markets (through dis- •• Relax the linear pricing rule (i.e., and simply crete sessions) is the way to go, while in the Unit- adapt to the fact there will be more than one ed States context the link between imbalances price-per-bidding zone and time period). and pricing (cost allocation) could be improved.

143 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

−− Balancing markets and products term solution is not to change the pricing rules Properly designing balancing markets is essential but to address the fundamental underlying mar- 4 to ensure that: 1) the market provides accurate ket structure that creates the market power (or incentives for flexibility and 2) all resources can to directly implement a regulated provision of effectively offer their flexibility potential to sys- the service). Marginal prices, of course, provide tem operators. sharper and more volatile prices, but these are the signals needed to reflect the real-time value Imbalance responsibility and imbalance settle- of flexibility and to incentivise resources that are ment are the two most important – and con- sensitive to these price signals. troversial – elements of designing a balancing mechanism: 4.1.2. Improving long-term signals without ruining short-term ones •• The definition of the balancing responsible parties (BRP) is a contentious issue when coupled The key message of this section is rather straight- with a dual imbalance pricing scheme. While forward: to the extent possible, renewable tech- dual imbalance pricing is applied, allowing gen- nologies should be exposed to the same market eration portfolios to BRP gives a competitive signals as conventional technologies. This is im- advantage to large companies and introduces portant to foster their efficient integration into entry barriers to small providers. the power system. While doing this, it must be •• Dual imbalance pricing does not exactly reflect guaranteed that the design of RES support mech- imbalance costs and therefore distorts the real- anisms is consistent with this guideline. Thus, the time price signal. economic incentives for RES should account for •• Today there is a growing trend to make RES the costs and the benefits derived from exposing increasingly responsible for their imbalances. RES generators to market signals. This is important to efficiently integrating larg- Long-term mechanisms, including convention- er volumes of RES in markets. al capacity mechanisms and specific renewable The definition of the balancing product can in- support schemes, are evolving due to the pen- troduce significant barriers to the participation of etration of renewable resources and distributed DER and RES. The following guidelines can help: generation. Generation adequacy mechanisms are at the top of the regulatory agenda in many •• Define innovative products to unlock the po- countries. It is a fact that conventional generation tential of new, flexible resources. suffers the increased short term price volatility •• Give different price signals to resources per- and the low energy prices when renewables such forming differently. as wind or solar enter the system in a short time, •• Separate the procurement of balancing energy, without giving time for the incumbent generation upwards reserves and downwards reserves. to adapt. However, the need for capacity remuner- •• To the extent possible, avoid limiting participa- ation is a topic mostly related to the presence of tion based on size or technology. caps on market prices to regulatory uncertainty as The pricing of balancing products can be based regards to the long term planning. As discussed, on either pay-as-cleared or pay-as-bid pricing. the deployment of variable generation has differ- In many cases, the procurement is bilateral and ent impacts in energy- and capacity-constrained therefore by definition the products are pay-as- power systems. In the former, RES technologies bid. In other cases, it has been implemented as do not complicate the operation of the system if pay-as-bid (and average reserve pricing charges), the transmission capacity does not create bot- with the objective of mitigating market power tlenecks, and RES can substantially contribute to and providing less volatile prices. In general, mar- generation adequacy. In capacity-constrained sys- ginal prices should set the market price. In case tems, non-dispatchable power plants do affect policy makers or competition authorities believe operations, and the participation of RES technol- that market power is a major problem, the long- ogies in capacity mechanisms is scarce and more difficult to quantify.

144 CONCLUSIONS AND RECOMMENDATIONS

−− Capacity markets design and RES propriate design, making them generally prefer- As far as technically possible, RES technologies able if the goal is market integration. should be allowed and expected to participate in When RES generation is distributed and is an in- generation adequacy mechanisms to the extent ternal component of some network users, the RES agents consider feasible and acceptable in same general considerations regarding support terms of risk (as is the case for any other tech- schemes hold. Specific implications for the de- nology). Nonetheless, when this participation is sign of support schemes and tariffs at distribu- accepted, it must be guaranteed that the compe- tion level are discussed in Section 4.2. tition in the generation-adequacy market is efficiently designed. In this context, some design elements are pivotal in achieving fair competition, 4.2. DISTRIBUTION NETWORKS AND namely constraints on tradable quantities and DISTRIBUTED ENERGY RESOURCES performance incentives (penalties and credits for As discussed in Chapter 3, the energy transition under- and over- performance). is driving profound changes in the “downstream Obviously, the participation of RES technologies segments” of the power sector i.e., the distribu- in capacity mechanisms, as in any other market tion and retail sectors. This is mainly driven by the segment, should be coherent with the support decentralisation of generation and the empower- being received by these technologies out of the ment of small- and medium-size consumers en- market. It may be possible that renewable re- abled by the diffusion of innovative technologies sources will still need economic incentives. How- and increased availability of information. What ever, their incentives should progressively be- follows is the summary of the recommendations come more market compatible in the future. From for promoting the large-scale deployment and ef- a market design perspective, support to renew- ficient integration of distributed energy resources able generation should provide optimal invest- whilst ensuring the reliability and economic via- ment signals while minimising market distortions. bility of the power system as a whole. Finally, support to prosumers and DER in gener- 4.2.1. Adapting regulation under high al should be harmonised with end-user tariffs to shares of distributed generation guarantee cost-recovery, as discussed below. Distribution networks are witnessing growing lev- −− RES support and wholesale market els of DG, driven by cost reductions and policies integration promoting the installation of low-carbon technol- There is a growing consensus on the need to in- ogies. Utilities may install DG units to supply en- tegrate RES in electricity markets. This requires ergy to their customers. Independent promoters the adoption of more market-compatible support or end consumers may invest in DG installations, schemes. However, it must be taken into account particularly in those countries where policies to that there is no perfect support scheme; all come promote RES favour smaller installations (e.g., with pros and cons: net metering, self-consumption or feed-in tariff policies). •• Production-based mechanisms are good at stimulating efficient investment decisions but The deployment of DG has positive effects for the may create short-term market distortions. system, for example in terms of reduction of net •• Capacity-based mechanisms are in gener- load (e.g., in the case of photovoltaic (PV) sys- al more market compatible than produc- tems production during peak demand) and trans- tion-based mechanisms, but they may be less mission losses (being the generation units close successful in stimulating optimal investment to consumption points). However, in the tradi- decisions. tional, passive approach to managing networks, the connection of high levels of DG may cause The risks inherent in capacity-based schemes an increase in network costs. This is particular- can be reduced via the implementation of an ap-

145 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

ly true when the network operator, either a utili- per investment”, policy makers ought to promote ty or an unbundled distribution company, has no pilot projects and facilitate the sharing of lessons 4 direct control over the location and operation of learnt: the generators, freely decided by DG investors or •• Useful measures include the creation of and end-users. With this in mind, regulators should participation in, national or international encourage these companies to adopt more active public-private collaboration networks for the strategies by establishing regulatory incentives promotion of innovative grid projects, the dif- that are well aligned with the desired transforma- fusion of best practices and the involvement of tion. Such regulatory changes become more rel- relevant stakeholders. The goal of these initia- evant as the penetration rate of DG grows and tives is to identify local priorities and promote should be progressively refined according to the knowledge-sharing to accelerate the process experience and data gathered by the network from testing to deployment. companies managing generation in their grids. •• Policy makers and regulators should set ad hoc Experience shows that in countries with low pen- financial incentives for utilities and distribution etrations of DG, utilities or distribution companies companies to invest in pilot projects in strategic that have scarce experience dealing with this phe- areas. Best practices include allocating these nomenon may see generation they do not direct- funds through competitive mechanisms and ly own as a threat to their activities. On the one introducing information disclosure obligations. hand, the installation of on-site generation can Subsequently, network operators should proceed lead to a decrease in their revenues due to low- to the deployment of proven solutions, leverag- er energy distributed. On the other hand, costs ing the experience gathered through demonstra- may remain constant or even increase when DG tion. To fully exploit the techno-economic effi- increases network capacity requirements. To pre- ciency potential of advanced grid solutions and vent network operators from opposing or delay- decentralised resources providing network ser- ing the connection of DG, regulators should fo- vices and flexibility, there should be transforma- cus on mitigating the negative impact DG may tion in the way regulators set revenues and incen- have distributed revenues and on network costs. tives for utilities and distribution companies. Where levels of DG penetration are low, it is recommended that regulators: A comprehensive reform of network regulations should be undertaken. Some of the recommend- •• Decouple the remuneration of network activ- ed regulatory provisions are: ities from the amount of energy consumed. Thus, any deviation between ex ante allowed •• Shift the regulatory focus from ensuring that revenues and the money collected through the distribution companies invest enough to as- tariffs can be corrected in subsequent periods. sessing their performance through measurable •• Compensate utilities and distribution compa- indicators (e.g., energy losses, customer satis- nies for the possible incremental costs driven faction, carbon emissions, peak load reduction, by DG, which are outside their control. This may etc.) – in other words, move from an input-ori- be done through revenue drivers or a pass- ented to an output-based regulation. through of costs. •• Remove biases towards capital expenditures so that utilities and distribution companies are The declining costs of RES will lead to a large- encouraged to solve network problems relying scale deployment of DG. In order to support and on the flexibility provided by DER, even if these integrate high amounts of DG efficiently, network avoid network reinforcements. operators should be incentivised to deploy ad- •• Perform cost assessments to foresee future in- vanced grid technologies, i.e., transition towards vestment needs instead of evaluating only past smarter distribution grids. Since innovative solu- information. Forecasts and justified business tions entail a higher risk for utilities and distribu- plans prepared by companies should become tion companies than conventional “iron and cop- key tools for regulators. •• Lengthen the duration of regulatory periods to

146 CONCLUSIONS AND RECOMMENDATIONS

allow companies to focus on delivering long- the existing network to host DG capacity and term efficiency rather than spending resources disclose this information, so that DG promoters on frequent and burdensome price reviews. may make better-informed applications. •• Introduce flexible remuneration formulas to ac- •• The previous recommendations can be rele- knowledge the existence of higher uncertain- vant even under low or moderate levels of DG ties, including profit-sharing schemes, menu penetration. Even in these initial stages, DG regulation and reopeners. capacity, albeit small in general terms, may be concentrated in specific areas, thus triggering 4.2.2. Encouraging distribution the need to adopt some of the previous solu- companies to act as system tions for managing locally high DG penetration operators in these areas. In response to the growing presence of new As DG levels become significant, network opera- types of distribution network users, utilities and tors can no longer follow a fit & forget approach distribution companies should actively interact to connect new generators. For more efficient with DER and make the most of their contribu- network development, they need to implement tion during the distribution network operations. advanced solutions for managing constraints This could help reduce grid reinforcement needs. close to real time, instead of relying exclusively This is relatively straightforward when these re- on grid reinforcements. In many cases, this will sources are owned and operated by a vertically imply resorting to the flexibilities of DER. Also, integrated utility. However, in the case of unbun- ad hoc regulatory mechanisms will be required dled utilities or distribution companies that need to enable utilities and distribution companies to to manage investor-owned DG units or resort to benefit from the potential of third-party flexibil- demand response resources, the mechanisms to ities. interact become more complex. •• Under moderate penetration levels, a suitable When this is the case, utilities and distribution solution would be to allow network operators companies need to become actual system oper- to offer DER non-firm connection agreements, ators, managing both the grid and the resources through which DG operators who decide to connected to it similar to the role conventionally opt in would pay lower network charges in played by TSOs/ISOs. In the case of integrated exchange for a possible limited curtailment of utilities, they would need to manage both their their production. Regulators should define the own and external resources, and regulatory mea- conditions under which distribution companies sures are required to prevent discriminatory be- may resort to this measure, the rules on how to haviour. set limits to the amount of RES production to The need for an enhanced co-ordination with ex- be curtailed and compensation mechanisms to ternal DG promoters starts from the moment that DG operators. new generators request a connection to the grid. •• With larger DG penetrations and developed de- In order to achieve a more cost-effective network mand response schemes, regulators should es- capacity allocation and better-informed connec- tablish more-advanced mechanisms to enable tion requests from promoters, the following pro- network operators to purchase services from visions are recommended: DER such as bilateral agreements and local markets. •• Network operators should manage grid con- •• Regulators should develop distribution grid nection applications for a given area in a more codes defining the technical and economic aggregated manner, instead of on a simple conditions under which these services may be first-come first-served basis. provided in a non-discriminatory way. •• Regulators should mandate utilities and distribution companies to assess the capability of

147 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

4.2.3. Enabling distribution sector, either directly or through another com- companies to act as neutral pany within the same group, regulators could be 4 market facilitators concerned about potential barriers for indepen- A consequence of the increasing decentralisation dent suppliers. of power systems is that the flexibility resources Alternative models for data management have connected to the distribution grid become pro- therefore been adopted in some countries, some- gressively needed to balance the systems’ gener- times with the creation of a new regulated entity ation and demand. Therefore, in contexts where responsible for data management. As these mod- distributed resources are not negligible anymore, els have their advantages and disadvantages, the regulation and market rules should enable the final choice depends upon the specific country participation of DER in system services through contexts and regulatory priorities. aggregation, if necessary. Regardless of the model chosen: Because DER are connected to the distribution grid, utilities and distribution companies have an Regulators should clearly define the roles, rights important role to play as mediators. This role is and obligations of different stakeholders, and particularly relevant in systems where a TSO/ISO the conditions under which the metering data is independent of the power utilities. Moreover, manager should provide access to retail market the existence of competitive energy and ancillary agents such as suppliers and aggregators: services markets represents an additional driv- •• Consumers should be the owners of the data er for the development of such business models. and have access to them to make better-in- Otherwise, TSOs/ISOs may rely exclusively on formed decisions on consumption and con- centralised resources. tracting. To unlock the flexibilities of DER and enable their •• Market agents should have access to meter- participation in upstream system services: ing data in transparent and non-discriminatory conditions once consumers’ permission has Regulators should develop new grid codes that been granted. define the: Finally, the deployment of innovative grid-edge •• Responsibility of utilities and distribution com- technologies may be facilitated through the par- panies as facilitators of DER participation in ticipation of utilities and distribution companies. upstream services and markets, which could These infrastructures comprises advanced me- include an ex ante technical validation and an tering technologies, distributed storage and elec- ex post verification of the provision. tric vehicle public charging facilities. Given that •• Information exchange required between utili- the ownership and operation of these infrastruc- ties or distribution companies and system/mar- tures may be subject to competition, key regu- ket operators. latory concerns arise in the case of unbundled Another factor that is driving changes in the con- distribution companies. For instance, when en- ventional role of utilities and distribution com- ergy storage is used both for grid support and panies is the liberalisation of retail markets. The to perform price arbitrage. Nonetheless, regula- reforms undertaken by some countries aim to tors may decide to allow these companies to own promote more-efficient electricity consumption these assets under certain conditions. by offering a wider choice in terms of suppliers •• Policy makers should promote the deployment and prices. In competitive retail markets, trans- of advanced metering technologies in those parent and non-discriminatory access to meter- countries pursuing the sustainable developing data is essential for a well-functioning market. ment of self-consumption, demand response Management of metering data is usually a task and distributed storage. for utilities or distribution companies. However, when these companies are also active in the retail

148 CONCLUSIONS AND RECOMMENDATIONS

•• Rollout strategies should be based on 4.2.4. Promoting sustainable cost-benefit analyses, potentially leading to development of self-consumption a mandated large-scale rollout that sets re- and demand response sponsibilities, time frames, technical require- The installation of DG units behind the consum- ments and economic conditions for cost-re- ers’ meters for self-consumption can facilitate covery. the growth of RES and yield important benefits •• Distribution companies have conventionally both for end users and for the power system. carried out this task and are a suitable can- Therefore: didate, especially for a large-scale rollout. Nonetheless, other entities may deploy these •• Policy makers and regulators should enable technologies; it may be more efficient, but and promote self-consumption, particularly for other barriers may arise in this case, such as small commercial and residential consumers. lack of standardisation or barriers to switch- •• Regulators should reduce to the extend possi- ing suppliers. ble administrative barriers to the installation of DG units for self-consumption, such as complex •• Regulators should enable distribution compa- permission procedures, financial guarantees, nies to influence the location of storage units costly technical connection studies, etc. within the grid to provide efficient grid support services. Meanwhile, traditional self-consumption and net metering practices may lead to a missing mon- •• This may be done either by introducing ex- ey problem that, if not properly addressed, would emptions on the unbundling rules, subject to hamper the recovery of the power system and size limitations or a preapproval by the reg- other energy policy costs when high shares of ulator or by enabling distribution companies generation are involved. to carry out competitive tenders to purchase grid services from independently owned Conventional tariff designs, i.e., those that are storage units. based mostly or even exclusively on a volumetric charge (e.g., USD/ kilowatt hour), coupled with •• Policy makers may intervene to kick-start the conventional energy meters, unable to record deployment of public charging infrastructure to consumption with a high time granularity, are at promote the use of EVs. Distribution companies the core of this problem. Several short-term reg- may be given this role, although policy makers ulatory measures have already been implement- may consider alternative providers in case this ed in different jurisdictions to tackle this problem, may collide with the unbundling rules. including:

•• Setting limits on the size of DG units used for self-consumption, implementing an overall cap on the amount of DG eligible for self-consumption at the national/local level or limiting the annual volume of energy consumption that can be compensated with energy surplus from DG units. •• Progressively moving away from purely volumetric tariffs and recovering part of the system’s fixed costs through a demand or a fixed charge, and adjusting the volumetric component accordingly. When introducing these changes in the rate structures, the impact on consumer bills should be carefully analysed. Fixed charges may jeopardise low-demand

149 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

consumers, debase demand response and en- each of these components are provided. Note ergy-efficiency efforts or encourage grid defec- that the degree of sophistication, i.e., the granu- 4 tion. This highlights the importance of ensuring larity of time and location dependent signals, can cost-reflective changes in the rate structure, as be modulated according to the penetration level advocated for in this report. of DER in a specific jurisdiction and the level of •• Implementing alternative schemes to compen- stress on the network. sate for the prosumers’ production surplus. This •• The value of energy should reflect the varying requires the progressive phasing out of net-me- cost of producing electricity over time, as well tering policies in favour of self-consumption as the variation in network effects by location. schemes with shorter netting intervals and a Time-of-use or dynamic energy prices would decoupling of the remuneration of electricity provide short-term economic signals promot- feed-in from the retail price. This energy sur- ing the efficient operation of DG units as well as plus can be valued, for instance, at the nodal demand-response and storage management. energy price or the wholesale market price. •• Network charges should reflect the impact of Even though the previous regulatory measures the end user on the utilisation of network as- help mitigate the missing money problem, they sets and grid expansion. Thus, network charges do not provide a real long-term solution for the should be based on the individual contribution efficient development of prosumers. This entails of the network user to the transmission and developing and implementing rigorous method- distribution costs, differentiated by time peri- ologies for calculating cost-reflective tariffs that od, location and whether power is being inject- promote efficient energy consumption and grid ed or withdrawn from the grid. Such charges utilisation whilst ensuring system cost-recovery. would provide end users with long-term eco- Such a retail tariff design enables the sustain- nomic signals promoting efficient location and able development of self-consumption and effi- investment decisions. Most important, they re- cient decisions regarding demand response and flect the actual contribution of network users to distributed storage. Whenever support measures network stress (peaks of injection or withdraw- are deemed necessary (e.g., to renewable gen- al), thus signalling the need for new network erators or certain consumer categories), these reinforcements. should be explicit and transparent in order to pre- •• The other regulated costs, when recouped vent undesired effects. through the electricity rates, should be allocated without distorting the economic sig- Cost-reflective tariffs, enabled by the deploy- nals sent by energy and network charges, ment of advanced metering technologies, would e.g., through a fixed charge per customer. This send efficient short- and long-term signals to end provision is particularly relevant in countries users reflecting their contribution to the system, where these costs account for a significant and ensuring the recovery of total system costs. share of total system costs, for instance, due Retail tariffs should be additive, i.e., calculated to support for RES. Nonetheless, should poli- considering separately the different cost compo- cy makers wish to avoid the undesired effects nents: 1) the value of electricity (energy) at each of fixed charges mentioned above, these costs time and location; 2) the cost of the transmis- may be recouped, at least partially, through sion and distribution grid and 3) other regulat- alternative revenue sources such as public ed costs. Hereafter, some guidelines to calculate budgets or taxes.

150 CONCLUSIONS AND RECOMMENDATIONS

151 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

GLOSSARY OF TERMS

Adequacy: existence of enough resources, either installed Distribution feeder: each of the medium-voltage (MV) lines or expected to be installed, to efficiently guarantee the that originate from primary distribution substations and sup- match of supply and demand in the long term. ply secondary substations or distribution transformers.

Ancillary services: services that are necessary to maintain Firmness: ability of resources already installed in the system the stability and security of power systems. to respond to actual requirements to meet existing demand efficiently. Balancing market: a market that is usually opened at gate closure, where agents submit upwards and downwards bids Firm energy and firm capacity: Energy and capacity that are through which the system operator will be able to fix po- expected to be available during some sort of scarcity condi- tential imbalances through least-cost resources; term used tions; products usually traded in the framework of adequacy in the European context. mechanisms.

Capacity mechanism: a regulatory instrument that reinforc- Gate closure time: moment at which the scheduled or fore- es the economic signal provided by other electricity markets casted production of generation units for a given period is (usually short-term ones) in order to attract enough invest- fixed for the purpose of determining the generation sched- ment and ensure system adequacy in liberalised power sec- ule. tors. Grid code: document that specifies the technical conditions Commercial losses: energy losses caused by non-technical that must be met by any agent ¬– whether a generator, con- aspects of the electricity supply, such as metering or billing sumer or grid operator – connected to the power system in errors, meter tampering, illegal connections, energy theft, order to ensure reliable, secure system operation. etc. Infra-marginal generation: all production units whose mar- Continuity of supply: a component of quality of service, de- ginal costs fall below those of the latest unit to be dispatched cided by the number and duration of supply interruptions in a certain time period following the cost-based merit order. experienced by electricity consumers. Infra-marginal rent: remuneration earned on a marginal Distributed energy resources (DER): small- to medium- market by infra-marginal generators. scale resources that are mainly connected to the lower volt- Variable generation: electricity production units whose outage levels (distribution grids) of the system or near the end put is not controllable or not continuously available, usually users, and that is comprised of three main elements: distrib- because of their reliance on a renewable primary energy re- uted generation, energy storage and demand response. source like wind or solar irradiation. Distributed generation (DG): a generating power plant serv- Levelised cost of electricity: the average cost of electricity ing customers on-site or providing support to a distribution per kilowatt hour (kWh) considering all fixed and variable network and connected to the grid at distribution-level volt- costs over the lifetime of an installation. ages. For the purpose of this report, the technologies used for distributed generation are renewable energy technolo- Merit order effect: the displacement of thermal generation gies. units from the generation economic dispatch during hours of high renewable production, driven by the much lower mar- also known as Demand-side response: demand-side man- ginal costs of RES technologies. agement or energy demand management, refers to the possibility to shift energy loads around in time. The manage- Operating reserves: generation capacity readily available to ment of small end-users must be automatically achieved at the system operator for solving contingencies. the user level, which requires online communications. In this Prosumers: electricity consumers that either show some regard, smart meters represent a key enabling technology of flexibility in their consumption as a response to price or vol- demand response. ume signals (proactive consumers), or that have generation connected on their side of the meter (producer-consumers). GLOSSARY OF TERMS

PICTURE CREDITS

Real-time market: a market similar to the balancing market, Page 10 © Shutterstock implemented by U.S. system operators to fix imbalances. Page 17 © Shutterstock Real-time pricing: retail tariff scheme in which the energy Page 18 © Shutterstock charge corresponds to the price of electricity on the wholesale market during a specific trading period, usually one Page 29 © Shutterstock hour. Page 31 © Shutterstock

Re-dispatch: process through which changes are made in the Page 32 © Shutterstock generation schedule due to changes in the supply or demand Shutterstock curves after a previously computed economic dispatch. Page 40 © Page 67 © Shutterstock Residual demand: net electricity demand resulting from subtracting the foreseen production of intermittent RES in a Page 90 © Shutterstock given period from the total system load in that period. Page 111 © Shutterstock Retailing (electricity): final sale of electrical energy to end Page 139 © Shutterstock consumers. Retailing constitutes the final segment of the electricity supply chain after generation, transmission, and Page 140 © Shutterstock distribution. Page 149 © Shutterstock System operator: institution in charge of managing the Page 151 © Shutterstock transmission network and, more generally, of guaranteeing the constant match between demand and supply in liber- Page 163 © Shutterstock alised power systems. Page 167 © Shutterstock Technical losses: energy losses caused by physical phenom- ena and dissipated in the form of noise or heat in the electricity network.

Time-of-use pricing: a retail tariff scheme in which the energy charge varies across blocks of hours (e.g., peak and off- peak hours) where the rate is pre-determined and constant over a period of, typically, several months.

Transmission system operator: an entity responsible for operating the transmission system in a given area and, where applicable, its interconnections with other systems, and for ensuring the long-term ability of the system to meet reasonable demands for the electricity transmission. ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

REFERENCES

ACER-CEER (2016a), Annual Report on the Results of Moni- AEEGSI (2012a), “Disposizioni Urgenti In Materia Di Preno- toring the Internal Electricity and Gas Markets in 2015, Agen- tazione Della Capacità Di Rete a Seguito Delle Ordinanze cy for the Cooperation of Energy Regulators – Council of 16 Maggio 2012, Nn. 1879, 1880, 1881, 1906, 1907, 1909, 1911, European Energy Regulators, Slovenia and Belgium. 1912, 1913, 1914, 1915, 1916 E 24 Maggio 2012, NN. Da 2004 a 2019, Del Consiglio Di Stato”, Deliberation 226/2012/R/EEL, ACER-CEER (2016b), Annual Report on the Results of Moni- Autorità per l’energia elettrica il gas e il sistema idrico, Milan, toring the Internal Electricity Markets in 2015, Agency for the 28 May 2012. Cooperation of Energy Regulators – Council of European Energy Regulators, Slovenia and Belgium. AEEGSI (2012b), “Disposizioni di Attuazione della De- liberazione dell’Autorità Per l’Energia Elettrica e il Gas ACER-CEER (2015), “Joint ACER-CEER response to the Eu- 226/2012/R/EEl, Relative alla Saturazione Virtuale delle Reti ropean Commission’s consultation on a new energy market Elettriche”, Deliberation 328/2012/R/EEL, Autorità per l’en- design”, Agency for the Cooperation of Energy Regulators ergia elettrica il gas e il sistema idrico, Milan, 26 July 2012. – Council of European Energy Regulators, Slovenia and Bel- gium. AEEGSI (2011), “Modifiche e integrazioni alla deliberazione dell’Autorità per l’energia elettrica e il gas ARG/elt 99/08, in ACER (2013), Capacity Remuneration Mechanisms and the materia di condizioni tecniche ed economiche per la connes- Internal Market for Electricity, Agency for the Cooperation of sione alle reti con obbligo di connessione di terzi degli impi- Energy Regulators, Slovenia, www.acer.europa.eu. anti di produzione (TICA), per la revisione degli strumenti al AEEG (2011), “Criteri e condizioni per la disciplina del siste- fine di superare il problema della saturazione virtuale delle ma di remunerazione della disponibilità di capacità produt- reti elettriche”, Deliberation ARG/ELT 187/11, Autorità per tiva di energia elettrica, ai sensi dell’articolo 2 del decreto l’energia elettrica il gas e il sistema idrico, 22 December 2011. legislativo 19 dicembre 2003, n. 379,”, Deliberazione 21 lu- AEEGSI (2010), “Modifications to the technical and eco- glio 2011 – ARG/elt 98/11, Autorità per l’Energia Elettrica e nomic conditions for the connection to electricity networks il Gas, Milan. with third-party access obligations of power generation AEEGSI (2015a), “Smart distribution system: Selective plants (original in Italian)”, Autorità per l’energia elettrica il promotion of innovative investments in the electricity dis- gas e il sistema idrico, Consultation Paper No. DCO 15/10, 25 tribution sector: Initial guidelines (original in Italian)”, Con- May 2010. sultation Paper No. 255/2015/R/EEL, Autorità per l’energia ANEEL (2012), “Resolução Normativa Nº 502, de 7 de Agos- elettrica il gas e il sistema idrico, Milan, 29 May 2015. to de 2012: Regulamenta sistemas de medição de energia AEEGSI (2015b), “Criteria for the definition of tariffs for elétrica de unidades consumidoras do Grupo B”, Agência the provision of the transmission and distribution services Nacional de Energia Elétrica, Brazil, 7 August 2012. and metering of electricity in the 5th regulatory period: Fi- Southern California Edison (2017), Southern California Edi- nal guidelines (original in Italian)”, Consultation Paper No. son’s Distributed Energy Resource Interconnection Map. Arc- 544/2015/R/EEL, Autorità per l’energia elettrica il gas e il GIS, California, available at: http://www.arcgis.com/home/ sistema idrico, Milan, 17 November 2015. webmap/viewer.html?webmap=e62dfa24128b4329bfc8b- AEEGSI (2015c), “Integrated text of the output-based reg- 27c4526f6b7. ulation of distribution and electricity metering for the reg- Barquín, J. (2014), “Spanish RES E Support History.” ENER- ulatory period 2016-2023 (original in Italian)”, Deliberation DAY 2014 - 9th Conference on Energy Economics and Tech- 646/2015/R/EEL, Autorità per l’energia elettrica il gas e il nology - A European Energy Market, Germany sistema idrico, Milan, 22 December 2015. Barroso, L. A. and C. Batlle (2011), “Review of support AEEGSI (2014), “A contribution to AGCOM survey M2M com- schemes for renewable energy sources in South America”, munication services,,” Memorandum 457/2014/I/com, http:// IAEE Third Quarter 2011, International Association for En- www.autorita.energia.it/allegati/inglese/457-14eng.pdf, ergy Economics, pp. 27-31. www.iaee.org/en/publications/ http://www.autorita.energia.it/allegati/inglese/457-14eng. newsletterdl.aspx?id=141. pdf.

154 REFERENCES

Batlle, C. (2013), “Chapter 7: Electricity generation and Burger, B. (2014), “Electricity production from solar and wholesale markets”, in Regulation of the Power Sector, wind in Germany in 2014”, Fraunhofer Institute for Solar En- I. J. Pérez -Arriaga (ed.), Springer-Verlag, London, ISBN 978- ergy Systems, ISE, Freiburg, Germany. www.ise.fraunhofer. 1-4471-5033-6. de/content/dam/ise/en/documents/publications/studies/ Electricity_Germany_2014_CW52.pdf. Batlle, C. and C. Ocaña (2013), “Chapter 3: Electricity regulation: Principles and institutions”, in Regulation of the Pow- Burger, S. (2015), “Utility business model innovation: A er Sector, I. J. Pérez-Arriaga (ed.), Springer-Verlag, London, techno-economic and strategic analysis of distributed ener- ISBN 978-1-4471-5033-6. gy systems”, Thesis submitted to the Technology and Policy Program, Massachusetts Institute of Technology, Boston. Batlle, C. et al. (2015), “The system adequacy problem: Lessons learned from the American continent”, in Capaci- Butler, L. and K. Neuhoff (2008), “Comparison of feed-in ty Mechanisms in the EU Energy Market: Law, Policy, and tariff, quota and auction mechanisms to support wind pow- Economics, L. Leigh Hancher, A. de Hauteclocque and M. er development”, Renewable Energy, Vol. 33/8, Elsevier, Sadowska (eds.), Oxford University Press, Oxford, United Amsterdam, pp. 1854-1867. Kingdom. CAISO (2017), “Using Renewables to Operate Low-Carbon BEE (German Renewable Energy Federation) (2014), Grid,” CAISO, California, US. http://www.caiso.com/Docu- “Factsheet: Renewables from Germany”, www.bee-ev.de/ ments/UsingRenewablesToOperateLow-CarbonGrid.pdf. fileadmin/Publikationen/Sonstiges/BEE_Factsheet_RE- CAISO, (2015), “Comments of the California Independent NEWABLES_FROM_GERMANY.pdf System Operator Corporation on Technical Workshops.” Benner, S. (2015), “A brief history of regulation signals Docket No. AD14-14-000. March 6, 2015. at PJM”, 9 June 2015, www.pjm.com/~/media/commit- California Solar Statistics (2016), California Solar Incen- tees-groups/committees/oc/20150701-rpi/20150701-item- tive (CSI) Program, Accessed on 7 March 2016, Available at 02-history-of-regulation-d.ashx. Source: https://www.californiasolarstatistics.ca.gov/. BEREC (2017), Expert workshop on Internet of Things, Brus- CEEPR (2005), Electricity Market Reform in the European sels 1 February 2017, http://berec.europa.eu/eng/events/ Union: Review of Progress toward Liberalization & Integra- berec_events_2017/151-berec-workshop-on-enabling-the- tion, MIT, USA internet-of-things CEER (2016), Key Support Elements of RES in Europe: Moving Bharatkumar, A. (2015), “Distribution network use-of-sys- towards Market Integration, CEER Report C15-SDE-49-03, tem charges under high penetration of distributed energy Council of European Energy Regulators, Brussels, 26 Jan- resources”, Thesis submitted to the Engineering Systems uary 2016. Division and Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, CEER (2015a), “CEER advice on customer data management Boston. for better retail market functioning: Electricity and gas”, Ref: C14-RMF-68-03, Council of European Energy Regulators, Bohn, R., M. Caramanis and F. Swcheppe (1984), “Optimal Brussels, 19 March 2015. pricing in electrical networks over space and time”, The RAND Journal of Economics, Vol. 15/3, Wiley, New Jersey, CEER (2015b), “The future role of DSOs: A CEER conclu- pp. 360-376. sions paper”, Ref: C15-DSO-16-03, Council of European En- ergy Regulators, Brussels, 13 July 2015. Bresler, F. S. (2014), “Energy and ancillary services uplift in PJM”, FERC Uplift Workshop Docket No. AD14-14, “Price for- CEER (2015c), “Status review of renewable and energy mation in energy and ancillary service markets operated by efficiency support schemes in Europe in 2012 and 2013”, regional transmission organisation and independent system Ref: C14-SDE-44-03, Council of European Energy Regula- operators”, 8 September 2014. tors, Brussels, 15 January 2015.

155 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

CEER (2014), “CEER status review on European regulatory EA Technology (2012), Assessing the Impact of Low Carbon approaches enabling smart grids solutions (“smart regula- Technologies on Great Britain’s Power Distribution Networks, tion”)”, Ref: C13-EQS-57-04, Council of European Energy Report No: 82530, Version: 3.1, prepared for Energy Net- Regulators, Brussels, 18 February 2014. works Association on behalf of Smart Grids Forum – Work Stream 3, 31 July 2012. CEER (2011), “CEER status review of regulatory approaches to smart electricity grids”, Ref: C11-EQS-45-04, Council of EDSO (2015), “Adapting distribution network tariffs to a European Energy Regulators, Brussels. decentralised energy future”, EDSO press release, European Distribution System Operators for Smart Grids, Brussels. CER (2014), “Integrated single electricity market (I-SEM), high level design for Ireland and Northern Ireland from EEI (2013), “Disruptive challenges: Financial implications 2016”, Draft Decision Paper, Commission for Energy Regu- and strategic responses to a changing retail electric busi- lation, Dublin, June 2014. ness”, Edison Electric Institute, Washington, DC.

CER (2005), “Group processing approach for renewable Eirgrid, SEMO, APX Power and SONI (2015), I-SEM Trialling generator connection applications: Connection and pricing of EUPHEMIA: Initial Phase Report, September 2015, www. rules: Direction to system operators”, CER/05/049, Com- sem-o.com/Publications/General/I-SEM%20Trialling%20 mission for Energy Regulation, Dublin, 6 April 2005. of%20EUPHEMIA%20-%20Initial%20Phase%20Report.pdf.

Cossent, R. et al. (2011), “Distribution network costs under ELIA (2016), Forecast and actual solar-PV power genera- different penetration levels of distributed generation”, Euro- tion, Solar PV Power Generator Data, ELIA 2016, Beligum, pean Transactions on Electrical Power, Vol. 21/6, Wiley On- http://www.elia.be/en/grid-data/power-generation/So- line Library, pp. 1869-1888. lar-power-generation-data/Graph

CPUC (2014), “Energy division staff proposal on residen- ENTSO-E (2014), “Supporting document for the network tial rate reform. staff proposal for residential rate reform in code on electricity balancing”, European Network of Trans- compliance with R.12-06-013 and assembly bill 327”, Califor- mission System Operators for Electricity, 6 August 2014, nia Public Utilities Commission, 3 January 2014. https://www.entsoe.eu.

CPUC (2013), “Decision adopting energy storage procure- EPIA (2014), Global Market Outlook for Photovoltaics 2014- ment framework and design program”, CPUC Decision 13- 2018, European Photovoltaic Industry Association, Belgium. 10-040, Date of Issuance 21 October 2013, California Public EPEX SPOT (2014), “Introduction of smart block bids – Utilities Commission. linked block orders and exclusive block orders”, Presenta- CPUC (2012a), “CPUC approves analog meter options for tion, European Power Exchange, January 2014, http://static. Edison and SDG&E customers who do not wish to have a epexspot.com/document/25495/Smart%20blocks%20 wireless smart meter”, CPUC Press Release, Docket #: A.11- -%20presentation. 07-020, California Public Utilities Commission, 19 April 2012.. EPRI (2016), Residential Off-Grid Solar + Storage Systems: CPUC (2012b), “CPUC approves analog meters for PG&E A Case Study Comparison of On-Grid and OffGrid Power for customers electing to opt out of smart meter service”, CPUC Residential Consumers, Report, Electric Power Research In- Press Release, Docket #: A.11-03-014,, California Public Util- stitute, Palo Alto, CA. ities Commission, 1 February 2012. ERCOT (2014), “Operating Reserve Demand Curve eLearn- DECC (2014), “Implementing electricity market reform ing”, ERCOT Market training, ORDC workshop, http://www. (EMR) – Finalised policy positions for implementation of ercot.com/services/training/course/107. EMR”, Policy document, Department of Energy & Climate Eurelectric (2015), “Prosumers – An integral part of the Change, London. power system and the market”, Paper, Eurelectric, Brussels. DECC (2012), “Feed-in Tariff with Contracts for Difference: Eurelectric (2013), “Active distribution system manage- Operational Framework (Annex A)”, Department of Energy ment: A key tool for the smooth integration of distributed & Climate Change, London, www.gov.uk generation”, Full discussion paper, Eurelectric, Brussels. DSIRE (2016), Find Policies & Incentives by State, Database Eurelectric (2012), Decentralised Storage: Impact on Future of State Incentives for Renewables & Efficiency, N.C. Clean Distribution Grids, Report, Eurelectric, Brussels. Energy Technology Center, N.C

156 REFERENCES

European Commission (2016a), “Proposal for a regulation FERC (2014a), “Notice of proceeding regarding the price of the European parliament and of the Council on the information in energy and ancillary services markets operat- ternal market for electricity (recast)”, COM(2016) 861 final, ed by regional transmission organisations and independent European Commission, Brussels. system operators”, Docket No. AD14-14-000, 19 June 2014, Federal Energy Regulatory Commission, Washington, DC. European Commission (2016b), “Proposal for a directive of the European parliament and of the Council on com- FERC (2014b), “Uplift in RTO and ISO markets, staff anal- mon rules for the internal market in electricity (recast)”, ysis”, Docket No. AD14-14-000, Federal Energy Regulatory COM(2016) 864 final, European Commission, Brussels. Commission, Washington, DC.

European Commission (2016c), “Proposal for a directive of FERC (2014c), “Coperator-initiated commitments in RTO the European parliament and of the Council on the promo- and ISO markets, staff analysis”, Docket No. AD14-14-000, tion of the use of energy from renewable sources (recast)”, Federal Energy Regulatory Commission, Washington, DC. COM(2016) 767 final, European Commission, Brussels. FERC (2014d), “Order on tariff filing and instituting Section European Commission (2015a), “Investment perspectives 206 proceeding”, Docket No. ER14 1050 000, issued on 30 in electricity markets (July 2015), Staff Working Document May 2014, Federal Energy Regulatory Commission, Wash- (2015)142, European Commission, Brussels. ington, DC.

European Commission (2015b), “Communication from the FERC (2011), “Frequency regulation compensation in the Commission to the European Parliament, the Council, the organised wholesale power markets”, Docket Nos. RM11-7- European Economic and Social Committee and the Com- 000 and AD-11-000, Order No. 755, issued on 20 October mittee of the Regions: Launching the public consultation 2011, Federal Energy Regulatory Commission, Washington, process on a new energy market design”, European Com- DC. mission, Brussels. FERC (2007), “Preventing undue discrimination and pref- European Commission (2015c), “Commission staff work- erence in transmission service”, Docket Nos. RM05-17-000 ing document: Best practices on renewable energy and RM05-25-000, Order No. 890, issued on 16 February self-consumption”, SWD (2015) 141 final, accompanying the 2007, Federal Energy Regulatory Commission, Washington, document COM (2015) 339 final, European Commission, DC. Brussels. Fraunhofer ISE (2016), Energy Charts, Fraunhofer ISE, Ger- European Commission (2014a), “Commission staff working many www.fraunhofer.de document: Cost-benefit analyses & state of play of smart Fraunhofer ISI (2015), “Negative Prices on the Electricity metering deployment in the EU-27”, SWD(2014) 189 final, Wholesale Market and Impacts of §24 EEG”, Fraunhofer In- European Commission, Brussels. stitute for Systems and Innovation Research ISI. 24 Decem- European Commission (2014b), Report from the Commis- ber 2015. sion: Benchmarking Smart Metering Deployment in the EU- Fraunhofer ISI (2014), (2014), Best Practice Design Fea- 27 with a Focus on Electricity, COM(2014) 356 final, Europe- tures for RES- E Support Schemes and Best Practice Meth- an Commission, Brussels. odologies to Determine Remuneration Levels, www.ecofys. European Commission (2014c), “Commission staff working com/files/files/diacore_best-practice-design-criteria-and- document: Country fiches for electricity smart metering”, lcoe_2014_final.pdf SWD(2014) 188 final, European Commission, Brussels. Galiteva, A. and K. Casey (2015), “Incentivizing flexibility European Commission (2009), “Directive 2009/72/EC con- – California perspective”, CAISO presentation at Germany cerning common rules for the internal market in electricity – United States Joint Expert Workshop “The future of elec- and repealing Directive 2003/54/EC”, European Commis- tricity market design”, Berlin, 18 June 2015. sion, Brussels Gestore mercati energetici (2017), Spot Electricity Market FERC (2016), “Settlement intervals and shortage pricing in (MPE), GME, Italy Available at: http://www.mercatoelettri- markets operated by regional transmission organisations co.org/en/mercati/mercatoelettrico/mpe.aspx and independent system operators”, Docket No. RM15-24- Gouveia, J. P. et al. (2014), “Effects of renewables penetra- 000, Order No. 825, issued on 16 June 2016, Federal Energy tion on the security of Portuguese electricity supply”, Ap- Regulatory Commission, Washington, DC. plied Energy, Vol. 123/15, Elsevier, Amsterdam, pp. 438-447.

157 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Graichen, P. (2015), “Flexibility in the power system cur- Huntington, S. et al. (2017), “Revisiting support policies for rent status and future challenges in Germany”, June 2015. RES-E adulthood: Towards market compatible schemes”, Presented by Dr. Patrick Graichen at “Future of Electricity Energy Policy, Vol. 104, Elsevier, Amsterdam, pp. 474-483. Market Design”, Germany–United States Joint Expert Work- (article in press at the time of this writing). shop, Federal Ministry for Economic Affairs and Energy IEA (2016), Re-powering Markets – Market Design and Reg- (BMWi), Berlin, Germany June 17-18, 2015. ulation during the Transition to Low-carbon Power Systems, Gribik, P. R., W. W. Hogan and S. L. Pope (2007), “Mar- Organisation for Economic Co-operation/International En- ket-clearing electricity prices and energy uplift”, Paper, Har- ergy Agency, Paris. vard Kennedy School, Cambridge, MA. IEA-PVPS (2016), Review and Analysis of PV self-consump- Herrero, I., P. Rodilla and C. Batlle (2015), “Electricity mar- tion policies, International Energy Agency, Paris. ket-clearing prices and investment incentives: The role of IFC (2013), “Russia’s new capacity-based renewable energy pricing rules”, Energy Economics, Vol. 47, Elsevier, Amster- support scheme: An analysis of decree No.449”, Internation- dam, pp. 42–51, doi:10.1016/j.eneco.2014.10.024. al Finance Corporation, Washington, DC, www.ifc.org. Hirth, L. and I. Ziegenhagen (2015), “Balancing power and Inter-American Development Bank (IADB) (2014), “Analy- variable renewables: Three links”, Renewable and Sustain- sis of the impact of increased non-conventional renewable able Energy Reviews, Vol. 50, Elsevier, Amsterdam, pp. 1035- energy generation on Latin American electric power sys- 1051, doi:10.1016/j.rser.2015.04.180. tems: Tools and methodologies for assessing future oper- Hoff, T.E., (2006). Photovoltaic Incentive Design Handbook ation, planning and expansion”, IDB Discussion Paper No. Photovoltaic Incentive Design Handbook. National Renew- IDB-DP-341, Inter-American Development Bank, Washing- able Energy Laboratory, Golden, Colorado, USA, (NREL/ ton, DC. SR-640-40845). December 2006. IRENA (2017a), Rethinking Energy 2017: Accelerating the Hogan, W. W. (2014), “Electricity market design and effi- global energy transformation. The International Renewable cient pricing: Applications for New England and beyond”, Energy Agency, Abu Dhabi. (forthcoming) Paper, Harvard Kennedy School, Cambridge, MA, 24 June IRENA (2017)b, Battery electricity storage, costs and mar- 2014, www.hks.harvard.edu/fs/whogan/Hogan_Pric- kets to 2030. The International Renewable Energy Agency, ing_062414r.pdf. Abu Dhabi. (forthcoming) Hogan, W. W. (2013), “Electricity scarcity pricing through IRENA (2015), Renewable Energy Auctions: A Guide to De- operating reserves”, Paper, Harvard Kennedy School, sign, The International Renewable Energy Agency, Abu Dha- Cambridge, MA, 25 April 2013, www.hks.harvard.edu/fs/ bi. whogan/Hogan_ORDC_042513.pdf. ISGAN (2014), “TSO-DSO interaction: An overview of cur- Hogan, W. W. (1995), “A wholesale pool spot market must rent interaction between transmission and distribution sys- be administered by the independent system operator: tem operators and an assessment of their cooperation in Avoiding the separation fallacy”, The Electricity Journal, Vol. Smart Grids”, ISGAN Discussion Paper, Annex 6, Power T&D 8/10, Elsevier, Amsterdam, pp. 26-37. Systems, Task 5, International Smart Grids Action Network, Hogan, W. W. and B. J. Ring (2003), “On minimum-uplift Paris. pricing for electricity markets”, Paper, Harvard Electricity Jamasb, T. and C. Marantes (2011), “Electricity distribution Policy Group, Harvard Kennedy School, Cambridge, MA. networks: Investment and regulation, and uncertain de- Hogan W. W. and ERCOT Staff (2013), “Back cast of interim mand”, Cambridge Working Papers in Economics CWPE 1115, solution b+ to improve real-time scarcity pricing”, white pa- University of Cambridge, UK. per, ERCOT, 21 March 2013, Harvard University. Jamasb, T. and M. Pollitt (2001), “Benchmarking and reg- Holttinen, H. et al. (2013), “The flexibility workout: Man- ulation: International electricity experience”, Utilities Policy, aging variable resources and assessing the need for pow- Vol. 9/3, Elsevier, Amsterdam, pp. 107-130. er system modification”, IEEE Power and Energy Magazine, Jendernalik, L. (2015), “Renewables integration: Challenges Vol. 11/6, Institute of Electrical and Electronics Engineers, and experiences from DSO perspective”, Renewable Energy New York, pp. 53-62. Training Week, Abu Dhabi, 28 January 2015.

158 REFERENCES

Jenkins, J. and I. Pérez-Arriaga (2014), “The remuneration MINETUR (2015), “Real Decreto 900/2015, de 9 de octubre, challenge: New solutions for the regulation of electricity por el que se regulan las condiciones administrativas, técni- distribution utilities under high penetrations of distributed cas y económicas de las modalidades de suministro de en- energy resources and smart grid technologies”, CEEPR WP ergía eléctrica con autoconsumo y de producción con auto- 2014-005MIT, MIT Center for Energy and Environmental Pol- consumo”, Ministerio de Industria Energía y Turismo. icy Research, Cambridge, MA. MISO (2015), “Comments of the midcontinent independent JRC (2017), European smart grid projects: number and system operator”, INC. Docket No. AD14-14-000, 6 March budget evolution, Joint research centre. Available at: http:// 2015. ses.jrc.ec.europa.eu/european-smart-grid-projects-num- MIT (2015), “The future of solar energy”, Massachusetts In- ber-and-budget-evolution stitute of Technology Energy Initiative, Cambridge, MA. JRC (2014), Smart Grid Projects Outlook 2014, Joint Re- Miura, J. P. (2014), “Integrated single electricity market search Centre of the European Commission, Luxembourg. (I-SEM): Energy trading arrangements proposed decisions”, Krad, I., E. Ibanez, and E. Ela (2015), “Quantifying the po- Stakeholder Forum on I-SEM Draft Decision, Dundalk, Ire- tential impacts of flexibility reserve on power system opera- land, 17 June 2014, www.allislandproject.org. tions..” In Green Technologies Conference (GreenTech), 2015 Moreno, B., A. J. López and M. T. García-Álvarez (2012), Seventh Annual IEEE (pp. 66-73). IEEE. “The electricity prices in the European Union: The role of Kraftwerk (2015), “Case study: Photovoltaic for commercial renewable energies and regulatory electric market reforms”, self-consumption”, Kraftwerk Renewable Power Solutions, Energy, Vol. 48/1, Elsevier, Amsterdam, pp. 307-313. http://www.kraftwerk-rps.com. Morgan, P. (2013), A Decade of Decoupling for US Energy Liu et al. (2015), “Extending the bidding format to promote Utilities: Rate Impacts, Designs, and Observations, Graceful demand response”, Energy Policy, Vol. 86, Elsevier, Amster- Systems LLC, Oregon, revised February 2013. dam, pp. 82-92. NERC (2015), “Glossary of terms used in NERC reliability Lo Schiavo, L. (2017)- ), “Competition and regulatory as- standards”, http://www.nerc.com/pa/Stand/Glossary%20 pects of electric vehicles charging: the Italian regulatory ex- of%20Terms/Glossary_of_Terms.pdf (updated 19 May 2015). perience,,” FSR workshop on “Competition And Regulatory Neuhoff K., et al. (2016), “Intraday Markets for Power: Dis- Aspects of Electric Vehicles”,, Florence, 13 January 2017. cretizing the Continuous Trading”, EPRG Working Paper, Lo Schiavo, L. (2015), “Regulation of smart distribution Cambridge systems in Italy: From input-based demonstration to out- Neuhoff, K. (2015), “Strengthening market price signals to put-based deployment”, CIRED, Lyon, France, 18 June 2015. encourage long-term investments in flexible assets”, pre- Mastropietro, P., P. Rodilla and C. Batlle (2015a), “Perfor- sented at Germany – United States Joint Expert Workshop mance incentives in capacity mechanisms: Conceptual con- “The future of electricity market design”, Berlin, 18 June siderations and empirical evidence”, working Economics of 2015. Energy and Environmental Policy, Vol. 6/1, IAEE, Cleveland, Neuhoff, K. and R. Boyd (2011), “International experiences pp. 149-163. of nodal pricing implementation”, Climate Policy Initiative, Mastropietro, P., P. Rodilla and C. Batlle (2015b), “National Berlin. capacity mechanisms in the European internal energy mar- Neuhoff K., et al. (2015), “Flexible short-term power trading: ket: Opening the doors to neighbours”, Energy Policy, Vol. Gathering experience in EU countries”, http://www.diw.de/ 82, Elsevier, Amsterdam, pp. 38-47. discussionpapers. Maurer, L. and L. Barroso (2011), Electricity Auctions: An New York DPS (2015), “Staff white paper on ratemaking and Overview of Efficient Practices,World Bank Publications, utility business models”, State of New York, Department of Washington, DC, ISBN 978 0 8213 8822 8. Public Service, CASE 14-M-0101, proceeding on motion of Miller, M. (2015), “Flexibility: How much? From where? And the commission in regard to reforming the energy vision, 28 how?” NREL presentation at Germany – United States Joint July 2015. Expert Workshop “The Future of Electricity Market Design”, Berlin, 18 June 2015.

159 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

New York DPS (2014), “Reforming the energy vision. NYS OFGEM (2013c), “Strategy decisions for the RIIO-ED1 elec- department of public service staff report and proposal”, tricity distribution price control: Tools for cost assessment: State of New York, Department of Public Service, CASE 14- Supplementary annex to RIIO-ED1 overview paper”, 26e/13, M-0101, proceeding on motion of the commission in regard 4 March 2013, Office of Gas and Electricity Markets, London. to reforming the energy vision, 24 April 2014. OFGEM (2013d), “Strategy decisions for the RIIO-ED1 elec- NYISO (2015), Post-Technical Workshop Comments of the tricity distribution price control: Uncertainty mechanisms: New York Independent System Operator, INC. Docket No. Supplementary annex to RIIO-ED1 overview paper”, 26c/13, AD14-14-000. March 6, 2015. 4 March 2013, Office of Gas and Electricity Markets, London.

Norther Power Grid (2017), Generation Availability Map, OFGEM (2010a), Handbook for Implementing the RIIO Mod- Northern Power Grid, UK https://www.northernpowergrid. el, Office of Gas and Electricity Markets, London. com/generation-availability-map OFGEM (2010b), “RIIO: A new way to regulate energy net- NREL (2016), Advanced Grid-Friendly Controls Demonstra- works: Final decision”, Office of Gas and Electricity Markets, tion Project for Utility-Scale PV Power Plants, Technical Re- London. port NREL/TP-5D00-65368, National Renewable Energy OFGEM (2009a), “Electricity distribution price control re- Laboratory, Colorado. view 5 final proposals – Incentives and obligations”, Office NREL (2013a), “The Western Wind and Solar Integration of Gas and Electricity Markets, London. Study Phase 2”, Technical Report NREL/TP-5500-55588, OFGEM (2009b), “Regulating energy networks for the fu- National Renewable Energy Laboratory, Colorado. ture: RPI-X@20 history of energy network regulation”, Of- NREL (2013b), “Solar energy and capacity value”, dissem- fice of Gas and Electricity Markets, London. ination brochure, National Renewable Energy Laboratory, OFGEM (2009c), “Regulating energy networks for the fu- Colorado. ture: RPI-X@20 principles, process and issues”, Office of Gas NREL (2011), “Solar renewable energy certificate (SREC) and Electricity Markets, London. markets: Status and trends”, National Renewable Energy Omel (2001), Electricity market activity rules, available at Laboratory, Colorado. http://www.omel.es/files/EMRules_2.pdf NREL (2008), “Determining the capacity value of wind: An Olmos L. et al. (2015), “D3.2 developments affecting the updated survey of methods and implementation”, Confer- design of short-term markets”, MARKET4RES, September ence Paper NREL/CP-500-43433, National Renewable En- 2015, market4res.eu/wp-content/uploads/D3.2_20151009_ ergy Laboratory, Colorado. final.pdf. NSGM (2017), “National Smart Grid Mission – Background”. O’Neill, R. P., T. Dautel and E. Krall (2011), “Recent ISO http://www.nsgm.gov.in/en/nsgm. software enhancements and future software and modeling OFGEM (2014), “RIIO-ED1: Final determinations for the plans”, www.ferc.gov/industries/electric/indus-act/rto/rto- slow-track electricity distribution companies: Business plan iso-soft-2011.pdf. expenditure assessment: Final decision”, Office of Gas and Papefthymiou, G., K. Grave and K. Dragoon (2014), Flexi- Electricity Markets, London. bility Options in Electricity Systems, Report, Project number OFGEM (2013a), “Strategy decisions for the RIIO-ED1 elec- POWDE14426, ECOFYS, Brussels. tricity distribution price control: Business plans and pro- Paulos, B. (2015), Regulating the Utility of the Future: Im- portionate treatment: Supplementary annex to RIIO-ED1 plications for the Grid Edge, GTM Research, Green Tech overview paper”, 26b/13, 4 March 2013, Office of Gas and Media, www.greentechmedia.com/research/report/regulat- Electricity Markets, London. ing-the-utility-of-the-future. OFGEM (2013b), “Strategy decisions for the RIIO-ED1 PCR-ESC (2015), “Euphemia performance, price coupling of electricity distribution price control: Financial issues: Sup- regions”, Presentation, European Stakeholder Committee, plementary annex to RIIO-ED1 overview paper”, 26d/13, 4 ENTSOE publications. March 2013, Office of Gas and Electricity Markets, London. PCR PXs, (2016), “EUPHEMIA Public Description: PCR Mar- ket Coupling Algorithm..” Version 1.3.

160 REFERENCES

Perez-Arriaga J. I., (2013), “Power system economics”, Reg- Reuter, W. H. et al. (2012), “Renewable energy investment: ulation of the power sector, Springer-Verlag London, London policy and market impacts”, Applied Energy, Vol. 97, Elsevi- er, Amsterdam, pp. 249-254. Perner, J. (2015), “Capacity markets in Europe”, US-Germa- ny Electricity Market Design Workshop, June 2015. Rivier, M., I. J. Pérez-Arriaga and L. Olmos (2013), “Elec- tricity transmission”, in Regulation of the Power Sector, PG&E (n.d.), “PG&E proposes smart energy reform in sup- Pérez-Arriaga, I. J. (ed.), Springer, London, ISBN 978-1-4471- port of growth of rooftop solar”, www.pge.com (accessed 5033-6. on 8 April 2015). RMI (2014), The Economics of Grid Defection: When and PJM (2015a), “Regulation Performance Impacts,,” presenta- Where Distributed Solar Generation Plus Storage Competes tion by Eric Hsia at PJM Market Implementation Committee, with Traditional Utility Service, Rocky Mountain Institute, October 7 2015. http://www.pjm.com/~/media/commit- CohnReznick Think Energy LLC and HOMER Energy LLC. tees-groups/committees/mic/20151007/20151007-item-08 a-mic-regulation-performance-impacts.ashx. RTE (2014), French Capacity Market, Report accompanying the draft rules, Réseau de Transport d’Électricité, Cedex, PJM (2015b), Comments of PJM Interconnection, L.L.C. France. Docket No. AD14-14-000. March 6, 2015. Scarf, H. E. (1994), “The allocation of resources in the pres- PJM (2014), “2017/2018 RPM base residual auction results”, ence of indivisibilities”, The Journal of Economic Perspec- PJM DOCS No. 794597, Pennsylvania New Jersey Maryland tives, Vol. 8/4, American Economic Association, Pittsburgh, Interconnection. PA, pp. 111-128. PJM (2013), “PJM Ancillary Services online training course”. SEPA (2015), “Utility solar market snapshot: Sustained Pennsylvania New Jersey Maryland Interconnection, www. growth in 2014”, Solar Electric Power Association, Washing- pjm.com/training/course-catalog.aspx. ton, DC, May 2015. PNM (2014), “Final Technology Performance Report: Smart Siewierski, T. (2015), Empirics of Intraday and Real-time Grid Demonstration Project: PV Plus Battery for Simultane- Markets in Europe: POLAND, Research report, http://hdl. ous Voltage Smoothing and Peak Shifting”, Public Service handle.net/10419/112779. Company of New Mexico. Smartgrid (2017), Smart Grid Demonstration Program. Pope, S.L., (2014), “Price formation in ISOs and RTOs, prin- Smartgrid.gov, USA ciples and improvements..” October 2014. FTI Consulting. Available at www.epsa.org. Smart Grids Task Force (2013), EG3 First Year Report: Op- tions on Handling Smart Grids Data, Smart Grids Task Force, Poudineh, R. and T. Jamasb (2014), “Distributed genera- Expert Group 3, European Commission, Brussels. tion, storage, demand response and energy efficiency as alternatives to grid capacity enhancement”, Energy Policy, Solar Power Europe (2015), Global Power Outlook for Solar Vol. 67, Elsevier, Amsterdam, pp. 222-231. Power 2015-2019, Solar Power Europe, Solar Power Europe, Brussels, Belgium Purkus, A. et al. (2015), “Market integration of renewable energies through direct marketing – Lessons learned from SP Energy Networks (2017), “Distributed generation heat the German market premium scheme”, Energy, Sustainability maps”, Scottish Power Energy Networks, Glasgow. Available and Society, Vol. 5/1, Springer, Berlin, pp. 1–13. at: https://www.spenergynetworks.co.uk/pages/connec- tion_opportunities.aspx Red Electrica de España (REE) (2016), Wind Generation Forecasting at REE. June 2016, http://www.ieawindfore- SP Energy Networks (2014), “SP distribution PLC: Use of casting.dk/-/media/Sites/IEA_task_36/BCN-2016/IEAW- system charging statement: Final notice: Effective from 1st indTask36-ExpGapsForecastingWorkshop_Barcelona_Ro- April 2014”, Scottish Power Energy Networks, Glasgow. driguez_2016.ashx?la=da Stoft, S. (2003), “The demand for operating reserves: RES-LEGAL (2015), “RES-LEGAL: Legal sources on renew- Key to price spikes and investment”, IEEE Transactions on able energy”, http://www.res-legal.eu (accessed on 1 July Power Systems, Vol. 18/2, pp. 470-477, doi: 10.1109/TP- 2015). WRS.2003.810679.

161 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

SWECO (2015), “Study on the effective integration of Dis- US DOE (2013), Grid Energy Storage, United States Depart- tributed Energy Resources for providing flexibility to the ment of Energy, Maryland. electricity system”. Final report to The European Commis- Vázquez, S., P. Rodilla, and C. Batlle (2014), Residual de- sion. 20 April 2015 mand models for strategic bidding in European power ex- TennetT (2017), “Market Review 2016”, Electricity market changes: Revisiting the methodology in the presence of a insights. TennetT. The Netherlands large penetration of renewables. Electric power systems research, Power Systems Research, Vol. 108, pp. 178-184. THINK Project (2013), From Distribution Networks to Smart Distribution Systems: Rethinking the Regulation of European Ventosa, M., P. Linares and I. J. Pérez-Arriaga (2013), “Pow- Electricity DSOs, European University Institute, Italy. er system economics”, in Regulation of the Power Sector, Pérez-Arriaga, I. J. (ed.), Springer, London, ISBN 978-1-4471- THINK Project (2012), Electricity Storage: How to Facilitate 5033-6. its Deployment and Operation in the EU’, European Universi- ty Institute, Italy. Van Vyve, M. (2011), “Linear prices for non-convex electricity markets: models and algorithms”, CORE Discussion Paper Trebolle, D. et al. (2010), “Distribution planning with reliabil- No. 2011050, Université catholique de Louvain, Center for ity options for distributed generation”, Electric Power Sys- Operations Research and Econometrics, Belgium. tems Research, Vol. 80/2, Elsevier, Amsterdam, pp. 222-229. Weißbach, T., and E. Welfonder (2009), “High frequency T&D World (2017), “India to Invest $44.9 Billion in Smart deviations within the European power system: Origins and Grid Infrastructure”, Smart Utility, T&D World, March 2017, proposals for improvement..” In Power Systems Conference Available at http://www.tdworld.com/smart-utility/india-in- and Exposition, 2009. PSCE’09. IEEE/PES (pp. 1-6). IEEE. vest-449-billion-smart-grid-infrastructure Xu, B. et al. (2016), “A comparison of policies on the partic- US Congress (2009), American Recovery and Reinvestment ipation of storage in U.S. frequency regulation markets”, in Act of 2009, Washington, DC, 6 January 2009, signed into proceedings of 2016 IEEE PES General Meeting. Boston, MA, Law on 17 February 2009. doi:10.1109/PESGM.2016.7741531.

162

ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

APPENDIX EUPHEMIA, THE PAN-EUROPEAN MARKET CLEARING ALGORITHM

The Price Coupling of Regions (PCR) is the ini- Using block orders and complex conditions to tiative of seven European Power Exchanges1 (at avoid being matched at a loss: this writing) to develop a single price coupling The constraints and cost structure of different solution to be used to calculate electricity prices resources can be bid in markets by using block across Europe, and allocate cross border capacity bids, for example: on a day-ahead basis. •• A thermal unit can use a combination of Under the PCR initiative, a common market clear- linked block orders to make sure it will recov- ing algorithm has been developed (EUPHEMIA, er its start-up cost and it will not produce less acronym of Pan-European Hybrid Electricity Mar- than its minimum power output. ket Integration Algorithm). This algorithm allows each Power Exhange to continue using the bid- •• A thermal plant can use the complex condi- ding formats previously in place in their respec- tion (where allowed) to ensure the recovery tive market, and opens the possibility of homog- of the start-up cost. enizing bidding formats across Europe in the •• By combining block orders, storage can sell future. during peak hours only if its off-peak pur- The basic bidding format in EUPHEMIA keeps on chase order has been accepted and buy/sell being the simple bid, though additional bidding combination is in the money (EPEX SPOT, formats that enable agents to hedge against un- 2014). certain market clearing outcomes are used, as •• Demand response may condition a load re- shown in the table A.1. duction in some periods, to an increased consumption in some others.

•• Plants can use exclusive orders to bid different production profiles in order to cover a wide range of possible market outcomes. In Box A.1 it is analyzed how, with the current bidding formats in Europe, it is quite complex to properly bid a resource that is called to play a key role in the future: storage plants.

164 See, for example, www.nordpoolspot.com APPENDIX

Table A.1 Bidding formats and description

Bid format Description

Simple orders

Buy or sell orders for a given volume and a limit price. It can be partially accepted if the Hourly step orders clearing price is equal to the bid price

Hourly linear Buy or sell order for a given volume and a pair of prices: an initial price at which the or- piecewise order der begins to be accepted and a final price at which the order is totally accepted

Block orders

Buy or sell order for a single price and volume and a period of consecutive hours that Regular block order can only be totally accepted

Regular block order that can be partially accepted, it includes a minimum acceptance Profile block order ratio condition

An Exclusive group is a set of block orders for which the sum of the accepted ratios Exclusive block orders cannot exceed 1. In the particular case of blocks that have a minimum acceptance ratio of 1 it means that at most one of the blocks of the exclusive group can be accepted.

Set of block orders where the acceptance of some blocks (children) is conditioned to the acceptance of others (parents)

Figure A.1 Linked block orders

MW Grandchild

Linked block orders Child

Parent

Trading Period

Source: Miura, 2014

A flexible “hourly” order is a block order with a fixed price limit, a fixed volume, and Flexible hourly block minimum acceptance ratio of 1, with duration of 1 hour. The hour is not defined by the order participant but will be determined by the algorithm (hence the name “flexible”)

Complex conditions

Series of hourly step orders conditioning to the recovery of a fixed quantity over the Minimum Income entire trading day

Limits the variation between the accepted volume at a period and the accepted volume Ramp limit at the adjacent periods

165 ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY

Box A1 The complexity of bidding some resources Currently, the I-SEM (the Integrated Single Electricity with existing formats: the case of storage Market, the electricity market of Ireland and Northern Ire- land) is performing a number of test over the EUPHEMIA None of the previous types of bids fits perfectly the char- algorithm to assess the ability of the bidding protocols acteristics and needs of storage, which is reducing the to perform complex optimal schedules such as those in- risk of being uneconomically committed throughout the volved by storage resources. The analysis of the results day subject to limited energy constraints. will provide valuable information about the suitability of Linked block orders seem to be the only alternative to current bidding protocols. At a high level this study con- partially hedge such risk: by combining (linking) two sists of two main phases: (i) first an initial Phase, involving block orders (one block for charging and another block trials performed by SEMO, and then (ii) the commercial for discharging), storage can sell during peak hours only phase, where trials will be performed by SEMO in con- if (i) its off-peak purchase order has been accepted and junction with industry participants. (ii) buy/sell combination is in the money (EPEX SPOT, At the time of this writing, the first phase has been com- 2014). However, this requires storage to anticipate the pleted and the report is already available (see Eirgrid hourly periods for consuming and the hourly periods for et al., 2015). The results show how “the current SEM algo- producing in advance. rithm is capable of running storage units with much great- The combined use of exclusive block orders with linked er flexibility throughout a day than EUPHEMIA due to the block orders could help mitigating this risk. This approach rigid profile of the linked block orders. The SEM clearing allows to enter multiple candidate profiles, created in line algorithm, a unit-commitment-like algorithm, has a num- with the principles outlined previously, each potential ber of features to optimally represent units with specific profile conforming an exclusive group. For example, the characteristics (e.g., pumped storage) while EUPHEMIA profile representing the limited running of the unit could has products which must be adapted, for the purpose of be entered in various different time periods (e.g., one pro- producing commitment starting point for dispatch, for a file during hours 10 – 13 and another profile during hours range of technologies.” 15 – 18). Nevertheless, the amount of blocks that can be In the figure A.2 it is schematically represented the pros used and combined is quite limited. and cons of the different alternative methods tested so far to bid storage units in EUPHEMIA.

Figure A.2 Characteristics of pumped hydro methodologies

Method Benefits Disbenefits

Accounts for reservoir levels; Inflexible due to limited hours; Linked Block (original) simplest linked block representation; does not restore to target level as used; units are price takers requires prediction of load and price

Accounts for reservoir levels; Inflexible as rejection of one hour causes Linked Block (extended) allows use in multiple time periods; rejection of all future hours; units are price takers requires prediction of load and price

Allows full utilisation; Does not account for reservoir level allows units to be price makers; limitations; Simple allows greatest flexibility to follow price would requires significant actions post signals; DAM; does not require prediction of load requires prediction of price

Source: Eirgrid et al., 2015

166 APPENDIX

IRENA HEADQUARTERS P.O. Box 236, Abu Dhabi United Arab Emirates

ADAPTING MARKET DESIGN TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY RENEWABLE VARIABLE OF SHARES TO HIGH DESIGN MARKET ADAPTING ADAPTING MARKET DESIGN www.irena.org TO HIGH SHARES OF VARIABLE RENEWABLE ENERGY Conclusions and Conclusions recommendations 4. networks Distribution resources and distributed 3. Wholesale design market 2. sector Power transition 1. Executive Summary 2017