The Full Cost of Electricity (FCe-)

Capacity Expansion and Dispatch Modeling: Model Documentation and Results for ERCOT Scenarios PART OF A SERIES OF WHITE PAPERS THE FULL COST OF ELECTRICITY is an interdisciplinary initiative of the Energy Institute of the University of Texas to identify and quantify the full-system cost of electric power generation and delivery – from the power plant to the wall socket. The purpose is to inform public policy discourse with comprehensive, rigorous and impartial analysis.

The generation of electric power and the infrastructure that delivers it is in the midst of dramatic and rapid change. Since 2000, declining renewable energy costs, stringent emissions standards, low-priced natural gas (post-2008), competitive electricity markets, and a host of technological innovations promise to forever change the landscape of an industry that has remained static for decades. Heightened awareness of newfound options available to consumers has injected yet another element to the policy debate surrounding these transformative changes, moving it beyond utility boardrooms and legislative hearing rooms to everyday living rooms.

The Full Cost of Electricity (FCe-) study employs a holistic approach to thoroughly examine the key factors affecting thetotal direct and indirect costs of generating and delivering electricity. As an interdisciplinary project, the FCe- synthesizes the expert analysis and different perspectives of faculty across the UT Austin campus, from engineering, economics, law, and policy. In addition to producing authoritative white papers that provide This paper is one in comprehensive assessment and analysis of various electric power a series of Full Cost system options, the study team developed online calculators that allow of Electricity white policymakers and other stakeholders, including the public, to estimate papers that examine the cost implications of potential policy actions. A framework of the particular aspects of research initiative, and a list of research participants and project sponsors the electricity system. are also available on the Energy Institute website: energy.utexas.edu

Other white papers All authors abide by the disclosure policies of the University of Texas at produced through the Austin. The University of Texas at Austin is committed to transparency and study can be accessed disclosure of all potential conflicts of interest. All UT investigators involved at the University of Texas with this research have filed their required financial disclosure forms with Energy Institute website: the university. Through this process the university has determined that energy.utexas.edu there are neither conflicts of interest nor the appearance of such conflicts. Capacity Expansion and Dispatch Modeling: Model Documentation and Results for ERCOT Scenarios

Neal Mann, Department of Mechanical Engineering, Cockrell School of Engineering Chen-Hao Tsai, Bureau of Economic Geology’s Center for Energy Economics, Jackson School of Geosciences Gürcan Gülen, Bureau of Economic Geology’s Center for Energy Economics, Jackson School of Geosciences Erich Schneider, Department of Mechanical Engineering, Cockrell School of Engineering Pedro Cuevas, McCombs School of Business Jim Dyer, McCombs School of Business John Butler, McCombs School of Business Tong Zhang, Department of Electrical and Computer Engineering, Cockrell School of Engineering Ross Baldick, Department of Electrical and Computer Engineering, Cockrell School of Engineering Thomas Deetjen, Department of Mechanical Engineering, Cockrell School of Engineering Rachel Morneau, Department of Mechanical Engineering, Cockrell School of Engineering

Mann, Neal, Tsai, Chen-Hao, Gülen, Gürcan, Schneider, Erich, Cuevas, Pedro, Dyer, Jim, Butler, John, Zhang, Tong, Baldick, Ross, Deetjen, Thomas, Morneau, Rachel, “Capacity Expansion and Dispatch Modeling: Model Documentation and Results for ERCOT Scenarios” White Paper UTEI/2017-4-14, 2017, available at http://energy.utexas.edu/the-full-cost-of-electricity-fce/.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | iii TABLE OF CONTENTS

List of Tables...... v List of Figures...... vi List of Acronyms...... vii Capacity Expansion and Dispatch Modeling...... 1 Long-Term Capacity Expansion Scenarios...... 20 Long-Term Capacity Expansion Results...... 29 2030 Hourly Dispatch Results...... 38 Conclusions and Future Work...... 41 References...... 42 Appendix A: ERCOT Plant Database for End-of-Year 2015...... 44 Appendix B: LCOE, LACE, and Net Value...... 53 Appendix C: Wind Rating Factors by Aggregate County (%)...... 56 Appendix D: ERCOT Hardwired Plant Additions for the CT Scenario...... 58 Appendix E: ERCOT Hardwired Plant Additions for the AR Scenario (in addition to the CT scenario)...... 59

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | iv LIST OF TABLES

Table 1 Key Differences across Four Models...... 10 Table 2 Resource Capacity by Fuel Type and Primary Mover...... 11 Table 3 Fuels Included in Modeling...... 12 Table 4 Wind Rating Factors by Load Zone (%)...... 14 Table 5 Distribution of Solar Resources in ERCOT...... 15 Table 6 Solar Rating Factors by Region (%)...... 16 Table 7 Capacities for Combined Heat and Power Plants in ERCOT...... 17 Table 8 Monthly Maximum Capacity Factors for Hydroelectric Generators in ERCOT...... 18 Table 9 Key Long-Term Capacity Expansion Scenario Assumptions...... 20 Table 10 Overnight Capital Expenditures by Generation Technology [$2015/kW], adapted from ERCOT (2015e)...... 23 Table 11 Financial Parameters for Calculating Base Capital Carrying Cost for Potential New Resources...... 25 Table 12 Levelized Investment Tax Credit for Solar PV Facilities [$/MW-Week)...... 25 Table 13 Retrofit Costs for Coal Power Plants ($/MW-Week)...... 26 Table 14 Retrofit Costs Assignment based on ERCOT (2011)...... 27 Table 15 Summary of Resources under the Two Scenarios (MW) with AURORAxmp...... 29 Table 16 Summary of Resources under the Two Scenarios (MW) with the Excel model...... 32 Table 17 VFC Rankings...... 36 Table 18 SCM 2030 Capacities under the Two Scenarios (GW)...... 36 Table 19 Comparison of 2030 Capacities from the Three Models (GW)...... 37

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | v LIST OF FIGURES

Figure 1 Example AURORAxmp System Diagram for ERCOT...... 3 Figure 2 Example Snapshot of an Hourly Flow...... 3 Figure 3 Merit order example with constant capacity...... 6 Figure 4 Merit order example with modified capacity...... 6 Figure 5 Approximate ERCOT Price Duration Curve – Top 6% of Hours...... 7 Figure 6 Screening Curve Method for New Technologies...... 9 Figure 7 Screening Curve Method Considering Existing Technologies...... 10 Figure 8 Composite Wind Counties. Composite counties in orange; source counties in yellow...... 12 Figure 9 ERCOT Load Zones...... 13 Figure 10 2014 Hourly Wind Profile for ERCOT North Load Zone...... 13 Figure 11 2014 Hourly Wind Profile for ERCOT South Load Zone...... 13 Figure 12 2014 Hourly Wind Profile for ERCOT West Load Zone...... 13 Figure 13 Wind Energy Production on Highest-Energy Day...... 14 Figure 14 Wind Energy Production on Lowest-Energy Day...... 14 Figure 15 Typical Winter Day Solar PV Production by Region...... 15 Figure 16 Typical Summer Day Solar PV Production by Region...... 15 Figure 17 Typical Winter Day Conglomerate Solar PV Production by Region...... 16 Figure 18 Typical Summer Day Conglomerate Solar Production by Region...... 16 Figure 19 Year-on-Year Load Growth Rates: Historical (2007–15), Forecast (2016–25); from ERCOT (2015d)...... 20 Figure 20 ERCOT Weather Zones (ERCOT 2016d)...... 21 Figure 21 Historical Energy Load Growth Rates by ERCOT Weather Zone, Calculated from ERCOT (2016e)...... 21 Figure 22 Historical and Forecast Energy Load Growth Rates by Weather Zone...... 22 Figure 23 Historical and Forecast Peak Load Growth Rates by Weather Zone...... 22 Figure 24 Annual Energy Demand Growth Rates in Four Main AURORAxmp Zones for ERCOT...... 22 Figure 25 Annual Decline Rates in Overnight Capital Expenditures of Wind, Solar, Battery Storage, CAES...... 24 Figure 26 Henry Hub Natural Gas Price Forecasts [$2015/MMBtu]...... 28 Figure 27 Coal Price Forecasts [$2015/MMBtu]...... 28 Figure 28 Annual Capacity Additions and Retirements with AURORAxmp...... 30 Figure 29 Annual Average Prices and Reserve Margins with AURORAxmp...... 31 Figure 30 Total Generation Output by Fuel Type from 2015 to 2030 with AURORAxmp...... 31 Figure 31 Total System Costs from 2015 to 2030 with AURORAxmp...... 31 Figure 32 Annual Capacity Additions and Retirements with the Excel model...... 33 Figure 33 Annual Average Prices and Reserve Margins with the Excel model...... 33 Figure 34 Total Generation Output by Fuel Type from 2015 to 2030 with the Excel model...... 34 Figure 35 Total System Costs from 2015 to 2030 with the Excel model...... 34 Figure 36 SCM: ERCOT 2030 CT Scenario without Existing Capacity...... 35 Figure 37 SCM: The Retirement of a Coal Unit. Left: Not retired. Right: Retired...... 35 Figure 38 SCM: ERCOT 2030 CT Scenario...... 36 Figure 39 SCM: ERCOT 2030 AR Scenario...... 36 Figure 40 Total Generation Output by Fuel Type in 2030 - Comparison of AURORAxmp, PLEXOS and Excel Results...... 38 Figure 41 Price Duration Curves for 2030 - Comparison of AURORAxmp and PLEXOS Results...... 39 Figure 42 Total System Cost in 2030...... 39

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | vi LIST OF ACRONYMS

AC Alternating current AEN Austin Energy load zone BATT Battery CAES Compressed air energy storage CCGT Combined cycle gas turbine CCS Carbon capture and storage CFE Comisión Federal de Electricidad, Mexico’s state-owned electric utility CHP Combined heat and power CPS CPS Energy load zone (San Antonio area) DC Direct current EIA U.S. Energy Information Administration EPA U.S. Environmental Protection Agency ERCOT Electric Reliability Cooperative of Texas FCe UT Energy Institute Full Cost of Electricity Study FERC Federal Energy Regulatory Commission FOM Fixed operating and maintenance costs FOR Forced outage rate IC Reciprocating internal combustion engine IGCC Integrated gasification combined cycle LACE Levelized avoided cost of electricity LCOE Levelized cost of electricity LCRA Lower Colorado River Authority load zone LFG Landfill gas LIG Lignite coal LNG Liquefied natural gas LT Long-term LTDEF Long-term Demand and Energy Forecast MIP Mixed-integer programming MISO Midcontinent Independent System Operator NG Natural gas NREL National Renewable Energy Laboratory O&M Operations and maintenance OCGT Open cycle gas turbine PUCT Public Utility Commission of Texas PV Solar photovoltaic RCEC Rayburn Country Electric Cooperative load zone RMT Resource modifier table SCM Screening curve method SPP Southwest Power Pool SRMC Short-run marginal cost ST Steam turbine SUB Subbituminous coal TCEQ Texas Commission on Environmental Quality VFC Variable fuel costs VOM Variable operating and maintenance costs WIND-C Wind in coastal county as defined by ERCOT WIND-O Offshore wind WT Wind turbine

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | vii 1| CAPACITY EXPANSION AND DISPATCH MODELING

The Full Cost of Electricity project (FCe) intends additional electricity that would be required if to capture and present costs associated with that power plant were not available (EIA 2016d). delivering customers one unit of electricity (kWh If LACE is greater than LCOE, that technology or MWh) by different generation technologies. can be considered competitive in that system.1 Typically, the levelized cost of electricity (LCOE) is used to provide a quick comparison across Electricity systems are complex and require real- technologies. The LCOE considers overnight time matching of demand and supply, which capital cost, operating and maintenance costs, comes with additional costs. In addition, load fuel costs (if applicable), average capacity factor growth and generation siting can require new of a typical plant for a given technology (i.e., investment in the transmission and distribution what percentage of the hours in a year the plant networks. In the meantime, there could be can generate at its nameplate capacity), and congestion on the transmission lines that could interest rate associated with capital cost. prevent the flow of electricity to certain regions, raising new costs. Sometimes, the cheapest Although the LCOE is commonly used and is units may not be fully dispatched. Independent valuable because of its simplicity, Rhodes et al. system operators (ISOs) manage these costs via (2017)—as part of the FCe project—demonstrate the ancillary services markets and, sometimes, that the standard treatment of LCOE ignores non-market payments. These and other practices regional differences in capital and operating by ISOs target grid reliability while meeting costs, availability of resources across locations demand in real-time at minimum cost and (e.g., sufficient amounts of wind speed or solar are known as security-constrained economic insolation), access to fuels infrastructure (e.g., dispatch and unit commitment. Neither LCOE natural gas pipelines, railways for coal delivery), nor LACE can fully capture these costs. geographic and electric power grid topography, and generation mix and load profiles in different Dispatch modeling does not capture all costs, grids, among other possible challenges. Rhodes either (e.g., cost of new transmission), but by et al. (2017) further improve upon the traditional construct, it approximates security-constrained LCOE calculations by incorporating costs of economic dispatch and unit commitment. As certain externalities including emissions of sulfur such, it follows economic dispatch subject to dioxide, nitrogen oxides, particulate matter, and price signals (including energy, capacity and greenhouse gases (carbon dioxide and methane). some ancillary services when applicable), operational characteristics of generation units, and Recently, the U.S. Energy Information transmission capacities across zones or nodes. Administration (EIA) has developed the levelized avoided cost of electricity (LACE) as a complement Dispatch modeling also allows users to investigate to LCOE. Rather than costs, LACE estimates multiple scenarios in terms of generation mixes the weighted average revenue that a certain as well as any other policies or sensitivities. technology would provide per unit of electricity In the long-term capacity expansion mode, in $/kWh like LCOE (Namovicz 2013). LACE dispatch models build and retire units based captures differences in generation portfolio on the economic viability of each individual mixes, grid topographies and load profiles across unit. The results provide information on total electricity systems. The same technology might capital expenditures; operating costs; fuel costs; have a different LACE value in different systems. emissions; revenues by unit; average, hourly, One interpretation of the LACE of a power plant is that it represents the cost to generate the 1 A detailed discussion of LCOE, LACE, and net value is provided in Appendix B.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 1 and regional prices; realized capacity factors tool, and an Excel-based model.2 Then, we employ over time; and reserve margins, among others. AURORAxmp, PLEXOS (Energy Exemplar Pty Ltd Although our models do not build or retire 2016), another commercial economic dispatch tool, transmission, users can investigate the need for and the Excel model for hourly dispatch simulations new transmission given the model results from for 2030 (8,760 hours). In these runs, we use the different scenarios (e.g., location of new builds 2030 generation portfolios from the long-term and price signals in different parts of a grid). capacity expansion simulation of AURORAxmp to compare results consistently. Finally, the In this way, users can evaluate whole system costs screening curve methodology (SCM) employs the from multiple dimensions. Moreover, assumptions 2030 demand forecast to estimate the minimum- on capacity factors, fuel, capital costs, operating cost generation portfolio to meet that demand. costs, and other characteristics can be changed over time if there are reasons to do so. In the For consistency across models, we employ the same calculation of LCOE, such inputs are typically assumptions to the extent possible: the portfolio treated as constant over the assumed life of a of existing units, cost structures for potential new typical plant of any technology. Such detailed resources, operational characteristics, and fuel model runs also improve LACE calculations prices, among others. However, the Excel model, since they provide the hourly wholesale price by its very nature, has simplifying assumptions estimates at different zones or nodes. and does not capture all the complexities of dispatch models. For example, generation In the hourly mode over one year, dispatch units are aggregated by fuel/technology type. modeling allows users to capture dispatch under Similarly, the SCM aggregates thermal generation various scenarios (e.g., weather perturbations units into technology groups while treating or unit outages). Users can also capture some wind and solar generation as negative load. of the ancillary services, the costs of which can be significant, especially during peak demand The Excel and SCM models are attractive because seasons. The incorporation of more variable their simplicity makes them user-friendly and resources, distributed energy resources, and storage creates the opportunity to implement them as can necessitate additional ancillary services and online tools. In this paper, we show that, using increase total cost of energy. The models cannot the same set of critical assumptions, all models calculate the cost of ancillary services that do not yield mostly similar results for long-term or currently exist. However, they can indicate when hourly simulations if some key assumptions are such additional services might be necessary to carefully captured. We can explain most of the balance the market and their potential value. discrepancies by inherent differences across models. We use two commercial software programs AURORAxmp for dispatch modeling in addition to two other approaches. In this paper, we describe In AURORAxmp, many of the input parameters these four models and present results from can be modified by users. These parameters two basic scenarios to demonstrate the include heat rates, ramp rates, start-up costs, importance of acknowledging system costs, the fixed and variable operating costs, minimum run versatility and usefulness of different modeling times, forced outages, scheduled maintenance tools, and shortcomings of each model. outages, and many more. These parameters are included while calculating economic BRIEF MODEL DESCRIPTIONS dispatch and long-term capacity expansion decisions. Some of the key parameters for the For long-term resource capacity expansion in units modeled are reported in Appendix A. ERCOT (2015 to 2030), we utilize AURORAxmp (EPIS LLC 2016), a commercial economic dispatch 2 The Excel model discussed in this paper was developed by FCe team members.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 2 AURORAXMP is flexible and allows the electricity into or exporting out of ERCOT, configuration of a wide scope and level of with a total of 820 MW capacity, although this granularity in both zonal and nodal network is relatively small when compared to installed representations. In this study, we utilize a zonal or operational capacity in ERCOT (Table 2). configuration with the transmission network and power flow as represented in Figure 1, which We use the mixed-integer programming (MIP) allows users to run scenarios via adjustment of algorithm with the objective function to maximize transmission capacities across eight zones. An the value of the resources being built and retired 3 example hourly flow is provided in Figure 2. for the long-term (LT) capacity expansion. Our (Graphics used with permission of EPIS LLC.) study horizon is 2015 through 2030. However, we expanded the LT study planning horizon to 2040 FIGURE 1 in order to have better simulation convergence in Example AURORAxmp System Diagram for ERCOT the later years (e.g., 2025 to 2030). This way, even the decision to build a new unit or retire an existing unit in 2030 is based on 11-year economics.4 FIGURE 2 Example Snapshot of an Hourly Flow

1 Houston;1 2 Houston;North; 3 South;2 North; 4 West;3 South; 12 4CPS; West; 13 12Austin CPS; Energy; 13 Austin 14 LCRA; Energy; 15 14Rayburn ; LCRA; 159999 Rayburn; represents 9999 effectively represents unconstrained effectively unconstrained flow. flow.

Transmission losses across Transmissionthese lines vary losses by 0.5% across to 1%; these the wheelinglines vary charge by is $0.66/MWh for all lines except for two DC connections0.5% to 1%; to Southwest the wheeling Power charge Pool (SPP).is $0.66/MWh Some of the transmission lines are capacity-constrained; for example,for all lines the capacityexcept for of thetwo line DC between connections zones 1 and 14 is 800 MW. This constraint can become significant if more generation is built in zone 14 and needs to be shipped to zone to Southwest Power Pool (SPP). SomeAURORAxmp of the assumes that new generators will be built and existing generators will be retired based on 1 (the greater Houston area), a growing load center where local air quality constraints limit generator transmission lines are capacity-constrained;economics. Thefor model’s forward-looking economic evaluation algorithm decides all new builds and permitting and construction. A similar constraint exists between zones 3 and 4, where much wind example, the capacity of the line betweenretirements zones and covers all future years through the final year of the planning horizon. The model capacity has been built and more installations are planned. There are two DC-tie resources for importing 1 and 14 is 800 MW. This constraintcalculates can become annual value 3(revenues In LT simulation, less cost)the mixed-integer over the programming planning horizon,(MIP) is formulated converts them to real values electricity into or exporting out of ERCOT, with a total of 820 MW capacity, although this is relatively significant if more generation is usingbuilt inan inflationzone 14 rate, andinternally calculates and passed the net to a presentthird-party solvervalue (MOSEK) (NPV) to using get a solution. a real discount rate. However, small when compared to installed or operational capacity in ERCOT (Table 2). AURORAxmp provides three optimization options: Traditional LT Logic, and needs to be shipped to zone 1new (the build greater and retirement decisionsMIP logic with are the madeobjective on function an annual to minimize basis. total system cost, and We use the mixed-integer programHouston (MIP) area), algorithm a growing with load the center objective where function local to maximizeMIP the logic value with the of objective function to maximize value (i.e., a mix of In each LT simulation iteration,resources the that model are most uses profitable). an updated We decided set to of employ new the resource MIP logic candidates and air quality constraints limit generator permitting3 the resources being built and retired for the long-term (LT) capacityretirement expansion. candidate Our sstudy to performwith horizon the objective the is standard to maximize chronological value because it provides commitment better stability and dispatch logic. The and construction. A similar constraint exists in energy-only markets such as the one in ERCOT. The developer of the 2015 through 2030. However, we expanded the LT study planningmodel horizon tracks to the 2040 resource in order costs to have and value of all new and existing resources based on the market prices better simulation convergencebetween in the zones later years3 and (e.g., 4, where 2025 muchto 2030 wind). This capacity way, even the decisionAURORAxmp, to EPIS Inc., concurred with this choice. For further details, developed in the iteration. seeThe www.epis.com. long-term logic with the MIP algorithm simultaneously makes retirement 4 build a new unit or retire anhas existing been builtunit in and 2030 more is based installations on 11and-year new are economics -planned.build decisions. with the objective function of maximizing the sum of the net present value of There are two DC-tie resources forresources importing on the system.4 ItEPIS makes recommends the build a minimum and retirementof five years of decisions extension beyond while the also adhering to user defined Figure 2 – Example Snapshot of an Hourly Flow study horizon. constraints which may include annual minimum or maximum builds, overall minimum or maximum resource additions, and retirement limits.

3 The model also includes extra constraints to limit the amount of change in system capacity that can In LT simulation, the mixed-integerThe Full program Cost of Electricity (MIP) is formulated (FCe-) internally and passed to a third-partyModel solver Documentation and Results for ERCOT Scenarios, July 2017 | 3 (MOSEK) to get a solution. AURORAxmp provides three optimization happenoptions: Traditionalbetween each LT Logic, iteration MIP logic, to with facilitate an optimal solution and to promote convergence. At the the objective function to minimize total system cost, and MIP logic withend the of objectiveeach iteration, function the to maximizeMIP logic value adjust s the current set of new builds and retirements. When the (i.e., a mix of resources that are most profitable). We decided to employsimulations the MIP logic converge, with the the objective model to will write the final Resource Modifier Table (RMT) with the new build maximize value because it provides better stability in energy-only marketsand retirement such as the decisionsone in ERCOT. to the The database. Convergence is deemed to have been met when the build developer of the AURORAxmp, EPIS Inc., concurred with this choice. For further details, see www.epis.com. decisions and resulting market prices have only changed within a tolerance chosen for the LT study from 4 EPIS recommends a minimum of five years of extension beyond the study horizon. one iteration to the next. 10

11

AURORAxmp assumes that new generators will upon previous work that modeled historical be built and existing generators will be retired wholesale electricity prices for the year 2011 based on economics. The model’s forward-looking (Garrison, 2014). PLEXOS uses a combination economic evaluation algorithm decides all new of linear programming and mixed-integer builds and retirements and covers all future years programming to find optimal solutions for through the final year of the planning horizon. unit commitment and economic dispatch. The The model calculates annual value (revenues Xpress-MP solver was used for all simulations. less cost) over the planning horizon, converts them to real values using an inflationPLEXOS rate, and The PLEXOS model simulates a single, hourly calculates the net present value (NPV)PLEXOS using is a commercial day-ahead energy market market modeling using only software the short-term that can be used to model power, gas, and a real discount rate. However, newwater build markets. and The PLEXOSschedule model optimization of the ERCOT module grid (ST(Texas Schedule). Interconnection) is based upon previous retirement decisions are made onwork an annualthat modeled basis. historicalThe ancillary wholesale services electr icitymarkets prices for frequencyfor the year 2011 (Garrison, 2014). PLEXOS uses regulation, spinning reserves and non-spinning a combination of linear programming and mixed-integer programming to find optimal solutions for unit In each LT simulation iteration, the model uses reserves were not modeled. The planning horizon commitment and economic dispatch. The Xpress-MP solver was used for all simulations. an updated set of new resource candidates and is set to one year with intervals of one hour retirement candidates to performThe the PLEXOS standard model simulate(8,760s hoursa single, total). hourly The day chronological-ahead market phase using is only the short-term schedule chronological commitment and optimizationdispatch logic. module (STset Schedule). to daily steps The to ancillary simulate services a day-ahead markets market. for frequency regulation, spinning The model tracks the resource costs and value of Additionally, the Transmission Detail parameter reserves and non-spinning reserves were not modeled. The planning horizon is set to one year with all new and existing resources based on the market is set to Nodal5 and the Heat Rate set to Detailed. intervals of one hour (8,760 hours total). The chronological phase is set to daily steps to simulate a day- prices developed in the iteration. The long-term 5 logic with the MIP algorithm simultaneouslyahead market. Additionally,The PLEXOSthe Transmission model uses Detail a subset parameter of the isparameters set to Nodal and the Heat Rate set to makes retirement and new-buildDetailed. decisions used by the AURORAxmp model. Some of the with the objective function of maximizing the specific parameters used are provided in Appendix The PLEXOS model uses a subset of the parameters used by the AURORAxmp model. Some of the sum of the net present value of resources on A. The ST Schedule module uses short-run specific parameters used are provided in Appendix A: ERCOT Plant Database for End-of-Year 2015. The the system. It makes the build and retirement marginal costs (SRMC) to determine the bids from decisions while also adhering toST user Schedule defined module useseach short generator-run marginal according costs to the(SRMC) following to determine equation: the bids from each generator constraints which may include annualaccording minimum to the followingor (1) equation: maximum builds, overall minimum or maximum resource additions, and retirement limits. (1)

The model also includes extra constraints to limit 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 = (𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 × 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝐻𝐻𝐻𝐻𝑎𝑎𝑎𝑎 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅) the amount of change in system Othercapacity possible that can short -runOther costs possible, including short-run+ grid(𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 service costs,𝑉𝑉𝑉𝑉𝑉𝑉 charges, 𝑂𝑂including&𝑀𝑀 𝐶𝐶𝐶𝐶𝐶𝐶 emissions 𝐶𝐶) costs, and heat production happen between each iteration, tovalues facilitate, are notan includedgrid in theservice model charges, and are emissions thus ignored. costs, and optimal solution and to promote convergence. At heat production values, are not included The transmission system is set up as a reduced zonal network as depicted in AURORAXMP is flexible and the end of each iteration, the MIP logic adjusts the in the model and are thus ignored. current set of new builds and retirements.allows the Whenconfiguration of a wide scope and level of granularity in both zonal and nodal network the simulations converge, the modelrepresentations. will write the In thisThe study, transmission we utilize asystem zonal isconfiguration set up as a reduced with the transmission network and power final Resource Modifier Table flow(RMT) as representedwith the new in zonalFigure network 1, which as allows depicted users in Figureto run scenarios1. It is solved via adjustment of transmission build and retirement decisions tocapacities the database. across eight usingzones a. AnDC example optimal hourlypower flow isapproximation provided in Figure 2Transmission losses across Convergence is deemed to have thesebeen metlines when vary by 0.5%using to 1%; a variable the wheeling shift factor charge method is $0.66/MWh and a for all lines except for two DC single slack bus. Line limits are enforced, while the build decisions and resultingconnections market prices to Southwest Power Pool (SPP). Some of the transmission lines are capacity-constrained; have only changed within a tolerance chosen for transformers, contingencies and losses are for example, the capacity of the line between zones 1 and 14 is 800 MW. This constraint can become the LT study from one iteration to the next. ignored. Unserved energy and dump energy (i.e., significant if more generationover-generation) is built in zoneare not 14 allowed and needs to ensure to be shippedthat to zone 1 (the greater Houston PLEXOS area), a growing load centerdemand where and generationlocal air quality are fully constraints balanced. limit generator permitting and construction. A similar constraint exists between zones 3 and 4, where much wind capacity has been PLEXOS is a commercial energybuilt market and modeling more installations are planned. There are two DC-tie resources for importing electricity into or software that can be used to model power, gas, exporting out of ERCOT,5 withNodal ahere total refers of to 820transmission MW networkcapacity, modeling although in PLEXOS this only. is relatively small when and water markets. The PLEXOS model of the Setting this parameter to Nodal preserves full detail of the transmission compared to installed or operational capacity in ERCOT (Table 2). ERCOT grid (Texas Interconnection) is based network and calculates optimal power flow. The other options, Regional and Zonal, calculate flows via a truck-route algorithm. We use the mixed-integer program (MIP) algorithm with the objective function to maximize the value of the resources being built and retired for the long-term (LT) capacity expansion. Our study horizon is 2015 through 2030. However, we expanded the LT study planning horizon to 2040 in order to have The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 4

5 Nodal here refers to transmission network modeling in PLEXOS only. Setting this parameter to Nodal preserves full detail of the transmission network and calculates optimal power flow. The other options, Regional and Zonal, calculate flows via a truck-route algorithm.

12 better simulation convergence in the later years (e.g., 2025 to 2030). This way, even the decision to build a new unit or retire an existing unit in 2030 is based on 11-year economics.

Figure 2.

Figure 1. It is solved usingExcel a DCModel optimal power flow approximation using a variableAlso unlike shift AURORAxmpfactor method and PLEXOS, and a single slack bus. Line limits are enforced, while transformers, contingenciespower and plants losses can abere dispatched ignored. and turned-off Unserved energy and dumpFor many energy purposes, (i.e., over including-generation) annual are hourly not allowed toinstantaneously ensure that demand in the Exceland model (i.e., there dispatch and long-term capacity expansion, it is generation are fully balanced. is no consideration of ramp rates). These start- possible to develop a simpler Excel-based model up costs are not considered in the economic Excel Model that can represent economic dispatch principles. dispatch rule implied by (2); however, they are For many purposes, includingOne simple annual approach hourly would dispatch be toand choose long- termthe capacityadded expansion, to the ittotal is possible system cost.to Further, there cheapest available facilities without exceeding develop a simpler Excel-based model that can represent economic dispatch isprinciples no consideration. One simple of minimum production their maximum capacity. Mathematically, this approach would be to choose the cheapest available facilities without exceedinglimits. their Demand maximum is satisfied hour by hour ignoring model can be expressed as the following: unit commitments in the previous or future capacity. Mathematically, this model can be expressed as the following: (2) hours and ignoring any minimum run times.

This myopic assumption may lead to unrealistic 𝑇𝑇 𝑁𝑁 (2) 𝑇𝑇 𝑁𝑁 solutions on(2 an) hour-by-hour basis. For example, 𝑚𝑚𝑚𝑚𝑚𝑚 𝑡𝑡 𝑛𝑛,𝑡𝑡 subject to 𝑛𝑛,𝑡𝑡 (∑ ∑ 𝑀𝑀𝑀𝑀𝑡𝑡 × 𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛,𝑡𝑡) the model may turn on and off a coal-fired power subject to AAG𝑛𝑛,𝑡𝑡 (∑𝑡𝑡=1 𝑛𝑛∑=1 𝑀𝑀𝑀𝑀 × 𝐴𝐴𝐴𝐴𝐴𝐴 ) subject toAAG 𝑡𝑡=1 𝑛𝑛=1 plant in consecutive hours, which is physically unlikely because these thermal generators require 𝑁𝑁 𝑁𝑁 several hours to safely ramp turbines up and down. In situations where coal is required to satisfy 𝑛𝑛,𝑡𝑡 𝑡𝑡 ∑ 𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛,𝑡𝑡 = 𝐷𝐷𝑡𝑡, ∀ 𝑡𝑡 𝑛𝑛∑=1 𝐴𝐴𝐴𝐴𝐴𝐴 = 𝐷𝐷 , ∀ 𝑡𝑡 demand, the model will tend to underestimate 𝑛𝑛=1 𝑛𝑛,𝑡𝑡 𝑛𝑛,𝑡𝑡 coal generation due to the technology switching 0 ≤ 𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛,𝑡𝑡 ≤ 𝑀𝑀𝑀𝑀𝑛𝑛,𝑡𝑡, ∀𝑡𝑡, ∀𝑛𝑛 where is the total numberwhere Nof is power the total plants, number is ofthe power total plants,number∀𝑡𝑡, ∀𝑛𝑛 T of is hoursoff in themore generation quickly than window reality. However, even with (8,760 hours for a year),the total market number price of hours in period in the, generation is the window total generationthis simplifying produced assumption, by power the model matches (8,760 hours𝑁𝑁 for a year), market price in period𝑇𝑇 , is the total generation produced by power 𝑁𝑁 (8,760 hours for a year), 𝑇𝑇MPt market price in period reality well on aggregate for 8,760 dispatch runs, plant during period , 𝑡𝑡 is the maximum generation available𝑛𝑛,𝑡𝑡 for power plant during period , 𝑀𝑀𝑃𝑃𝑡𝑡 𝑡𝑡 𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛,𝑡𝑡 t, AAG𝑀𝑀𝑃𝑃 n,t is the total generation 𝑡𝑡produced𝐴𝐴𝐴𝐴𝐴𝐴 by power and the assumption may not preclude near-optimal and is the demand during𝑛𝑛 ,𝑡𝑡period . Typically, is the cost of the most expensive dispatched 𝑛𝑛 𝑡𝑡plant𝑀𝑀𝑀𝑀 n𝑛𝑛, 𝑡𝑡during period t, MG is the maximum solutions𝑛𝑛 for long-term capacity𝑡𝑡 expansion. technology𝑛𝑛 , but we do 𝑡𝑡 account𝑀𝑀𝑀𝑀 for price spikes in n,tthe ERCOT market as we describe below.𝑛𝑛 𝑡𝑡 technology𝑡𝑡 , but we dogeneration account for available price spikes for power in the 𝑡𝑡plant ERCOT n during market as we describe below. 𝐷𝐷𝑡𝑡 𝑡𝑡 𝑀𝑀𝑃𝑃𝑡𝑡 period t , and D is the demand during period t. The dispatch problem is solved by ordering all The second constraint limits the generationt to the maximum capacity available for each power plant. Typically, MP is the cost of the most expensive available power plants from the cheapest to the For base load power plants, we adjustt the maximum capacity on a percentagemost basis expensive by technology variable and costs 6 and then deploying by month to approximatedispatched maintenance technology, periods. but Forwe doexample, account we for reduc priceed the maximum capacity of all by month to approximatespikes maintenance in the ERCOT periods. market For as example, we describe we below.reduced thethem maximum in merit capacity order to of satisfy all demand at any given existing coal plants by 20% in March, April, October, November and Decemberhour. to Thisreflect approach historical assumes that the relative maintenance activities.The second constraint limits the generation to the efficiencies between plants are fixed within a year maximum capacity available for each power plant. in the model, although they can vary from year to Unlike the eight zonesFor in AURORAxmp base load power and plants, PLEXOS, we the adjust Excel the model maximum treats ERCOTyear. Likewise, as a single because node. fuel prices are correlated Accordingly, the modelcapacity implicitly on assumesa percentage that basisevery by marke technologyt participant seesin the the same short marketrun, they price are, changed annually in the that there are no transmissionand by month constraints, to approximate and that maintenancedifferences in load growthExcel across model the forecasts, zones are but not not hour to hour. If the captured. periods. For example, we reduced the maximum order of the marginal, or variable costs, is known, capacity of all existing coal plants by 20% in the problem of determining which power plants Also unlike AURORAxmpMarch, and PLEXOS,April, October, power Novemberplants can andbe dispatched December and turnedshould-off be instdispatchedantaneously in any in given hour is trivial. the Excel model (i.e., thereto reflect is no historicalconsideration maintenance of ramp activities.rates). These start-upThis costs methodology are not considered is known in as merit order dispatch. the economic dispatch rule implied by (2); however, they are added to the total system cost. Further, Unlike the eight zones in AURORAxmp and To clarify this proposed solution methodology, there is no consideration of minimum production limits. Demand is satisfied hour by hour ignoring unit commitments in the previousPLEXOS, or the future Excel hours model and treats ignoring ERCOT any asminimum a runconsider times. a simple example of a market with five commitments in the previoussingle node. or future Accordingly, hours and the ignoring model implicitly any minimum runpower times. plants, each with a capacity of 100 MW: PP1, assumes that every market participant sees PP2, PP3, PP4 and PP5, where their marginal prices 13 the same market price, that there are no (MPn,t) are $10, $40, $70, $12013 and $160/MWh transmission constraints, and that differences in load growth across the zones are not captured. 6 Variable costs in the Excel model include variable O&M costs plus fuel costs, which is the same as equation (1) used in PLEXOS.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 5 This myopic assumption may lead to unrealistic solutions on an hour-by-hour basis. For example, the model may turn on and off a coal-fired power plant in consecutive hours, which is physically unlikely because these thermal generators require several hours to safely ramp turbines up and down. In situations where coal is required to satisfy demand, the model will tend to underestimate coal generation due to the technology switching off more quickly than reality. However, even with this simplifying assumption, the model matches reality well on aggregate for 8,760 dispatch runs, and the assumption may not preclude near-optimal solutions for long-term capacity expansion.

The dispatch problem is solved by ordering all available power plants from the cheapest to the most expensive variable costs6 and then deploying them in merit order to satisfy demand at any given hour. This approach assumes that the relative efficiencies between plants are fixed within a year in the model, although they can vary from year to year. Likewise, because fuel prices are correlated in the short run, they are changed annually in the Excel model forecasts, but not hour to hour. If the order of the marginal, or variable costs, is known, the problem of determining which power plants should be dispatched in any given hour is trivial. This methodology is known as merit order dispatch.

To clarify this proposed solution methodology, consider a simple example of a market with five power plants, each with a capacity of 100 MW: PP1, PP2, PP3, PP4 and PP5, where their marginal prices ( ) are $10, $40, $70, $120 and $160/MWh respectively (Figure 3). The addition of incrementalwith the same power demand of 250 MW, the market price would be $120/MWh, with generation 𝐷𝐷𝐷𝐷𝑛𝑛,𝑡𝑡 plants as we move right on the x-axis gives shape to the supply curve. If the supplyprofile curve does not , and . change over time, the only variable that sets the market price is the quantity demanded.Repeating If one this hour process of for the 8,760 hours over a year results in the annual observed generation mix and 𝐴𝐴𝐴𝐴𝐴𝐴𝑃𝑃𝑃𝑃1,1 = 50 𝑀𝑀𝑀𝑀, 𝐴𝐴𝐴𝐴𝐴𝐴𝑃𝑃𝑃𝑃2,1 = 50 𝑀𝑀𝑀𝑀 𝐴𝐴𝐴𝐴𝐴𝐴𝑃𝑃𝑃𝑃3,1 = 100 𝑀𝑀𝑀𝑀 𝐴𝐴𝐴𝐴𝐴𝐴𝑃𝑃𝑃𝑃4,1 = 50 𝑀𝑀𝑀𝑀 demand is 250 MW, the market price, , would be $70/MWh, with a generationsystem profile costs given the available capacity of each technology and hourly demand levels. of FIGURE 3 and . FIGURE 4 𝑀𝑀𝑃𝑃𝑡𝑡 𝑃𝑃𝑃𝑃1,1 Merit order𝑃𝑃𝑃𝑃 example2,1 with constant capacity.𝑃𝑃𝑃𝑃3,1 Merit order example with modified capacity. 𝐴𝐴𝐴𝐴𝐴𝐴 = 100 𝑀𝑀𝑀𝑀, 𝐴𝐴𝐴𝐴𝐴𝐴Figure 3= – Merit100 𝑀𝑀𝑀𝑀order example 𝐴𝐴𝐴𝐴𝐴𝐴 with constant= 50 capacity. 𝑀𝑀𝑀𝑀 Figure 4 – Merit order example with modified capacity.

As an alternative, suppose that power plants 1 and 2 (PP1 and PP2) only have capacitiesConsidering of 50 theMW number each, of power plants and the fast processing objective for the Excel model, ERCOT but the dispatch pricesrespectively for all five plants (Figure remain 3). unchanged The additionfrom the previous of incremental example.power Theplants dispatch areLCOE, aggregated net into value, different and groups costs, that simplifying are each managed the as model one modular unit. These order for the five plantspower will remainplants the as same, we movebut the supplyright curve on the will shiftx-axis to the groupsgives left (Fi aregure formed 4).runs Now considering, for both whether capacity the plants expansion were pre-existing and 8,760 or constructed dispatch during the capacity shape to the supply curve. If the supply curveplanning does horizonanalyses., as well as 7the The fuel typeexisting and heat technology rate. Each group categories of generator useds will be referred to as a 6 Variable costs in the Excel model include variable O&M costs plus fuel costs, which is thetechnolog same as equationy. The fact (1) that each technology group is analyzed as one unit means that all individual units used in PLEXOS. not change over time, the only variable that setswill share the the samein the LACE, ERCOT LCOE, net modelvalue, and for costs 2015, simplifying are defined the model runsbelow: for both capacity market price is the quantity demanded. If oneexpansion hour and 8,760 dispatch analyses.7 The existing technology categories used in the ERCOT model for • 14 of demand is 250 MW, the market price, MP2015, would are defined below:Existing Coastal Wind: Coastal & offshore t wind generation (13 wind farms, 1,845.4 MW) be $70/MWh, with a generation profile of  Existing Coastal Wind: Coastal & offshore wind generation (13 wind farms, 1,845.4 MW) AAG = 100 MW, AAG = 100 MW and  Existing Inland Wind: Onshore wind generation (137 wind farms, 14,041.7 MW) PP1,1 PP2,1  • Existing Inland Wind: Onshore wind AAG = 50 MW. Existing Solar Photovoltaic (PV): All utility-scale PV generation (15 power plants, 287.7 MW) PP3,1  Hydro: All hydrogeneration-electric generation (137 wind (29 power farms, plants 14,041.7, 555.1 MW MW))  Biomass: All biomass generation (7 power plants, 243.5 MW) As an alternative, suppose that power plants 1 and  Nuclear• : AllExisting nuclear generation Solar (4Photovoltaic power plants, 5,133 (PV) MW:) All utility- 2 (PP1 and PP2) only have capacities of 50 MW Existing Lignite:scale A llPV lignite generation coal-fueled generation (15 power (12 power plants, plant 287.7s, 7,142.0 MW) MW) each, but the dispatch prices for all five plants  Existing Bituminous: All bituminous coal-fueled generation (21 power plants, 12,637.0 MW) remain unchanged from the previous example.  Existing• NonHydro-Cycling: GasAll: Allhydro-electric non-cycling natural generation gas fueled generation (138 power plants, 47,762.4 MW) The dispatch order for the five plants will remain  Existing Cycling(29 Gas:power All cycling plants, natural 555.1 gas fueled MW) generation (56 power plants, 3,789.2 MW) the same, but the supply curve will shift to the left • Biomass: All biomass generation (Figure 4). Now, with the same demand of 250 MW, the market price would be $120/MWh,7 withA detailed discussion(7 of powerLCOE, LACE plants, and net value243.5 is provided MW) in Appendix B: LCOE, LACE, and Net Value. generation profileAAG = 50 MW, AAG = 15 PP1,1 PP2,1 • Nuclear: All nuclear generation (4 50 MW, AAG = 100 MW and AAG = 50 MW. PP3,1 PP4,1 power plants, 5,133 MW) Repeating this process for the 8,760 hours over a year results in the annual observed generation • Existing Lignite: All lignite coal-fueled mix and system costs given the available capacity generation (12 power plants, 7,142.0 MW) of each technology and hourly demand levels. • Existing Bituminous: All bituminous coal-fueled Considering the number of power plants and generation (21 power plants, 12,637.0 MW) the fast processing objective for the Excel model, ERCOT power plants are aggregated into different • Existing Non-Cycling Gas: All non- groups that are each managed as one modular unit. cycling natural gas fueled generation These groups are formed considering whether the (138 power plants, 47,762.4 MW) plants were pre-existing or constructed during • Existing Cycling Gas: All cycling natural gas the capacity planning horizon, as well as the fuel fueled generation (56 power plants, 3,789.2 MW) type and heat rate. Each group of generators will be referred to as a technology. The fact that each technology group is analyzed as one unit means 7 A detailed discussion of LCOE, LACE and net value is provided in that all individual units will share the same LACE, Appendix B.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 6 The Excel model can be utilized as a dispatch the more complex dispatch algorithms may look model by simply ignoring new technologies at the revenues and costs over the forecast life of beyond the assumed hardwired additions. the plant (or through the last year of model run) In this mode, the results are analogous to all when making expansion and retirement decisions. other models where the objective is simply to provide the cheapest electricity possible When considering expansion decisions past 2015, given the existing technologies (e.g., minimize the Excel model also considered adding six new total system cost while meeting demand). technology categories designed to represent the likely characteristics of new builds: New Coastal In the long-term capacity expansion mode, this Wind, New Inland Wind, New PV, New Coal, model starts from the available capacity and New Cycling Gas, New Non-Cycling Gas. The adds plants to the system from the next most Excel model also includes a “Big M” technology “valuable” technology to meet growing demand to ensure that demand would be satisfied in every or to replace units that have been retired in hour of the year. The cost of Big M generation previous periods. Expansion and contraction is not added to the costs reported by the model, occur in steps that represent an average power but it provides a signal to the model of the value plant capacity in a technology group. For example, of the last needed units of generation. Further, coal adjustments are in 500 MW increments. large levels of unserved generation (e.g., high levels of electricity provided by Big M) indicate The difference between LACE and LCOE, that the electricity system may be in peril: it is not referred to as net value, accounts for revenues economic to serve unfilled load. The marginal and cost, making it possible to estimate the profit cost of Big M was set to be $1 more than the most of each technology in each year. The net value expensive available technology so that it would is used to calculate an ordered list of plants that always dispatch last. During the runs of the model, are candidates to be added to the system. This the percentage of load that was provided by Big merit order is recalculated every year because of M was about 6%, which was treated as additional changes in demand, capital costs, operating and gas generation in the results presented below. maintenance costs, subsidies, and fuel costs. It is important to note that these expansion decisions in The Excel model does not include random events the Excel model are myopic in that they only look that might influence price like unexpected outages at the revenue and cost for a given year. In contrast, or extreme weather events. However, capturing

FIGURE 5

Approximate ERCOT Price DurationFigure Curve 5 –– TopApproximate 6% of Hours ERCOT Price Duration Curve – Top 6% of Hours

Every time a new megawatt of generation capacity is added, the capacity factor for all other The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 7 technologies drops because the new project’s generation will displace other power plants, assuming | demand is constant and there are no retirements. Therefore, for every new megawatt constructed, LACE for all technologies will drop until there are no more profitable projects left to add. This point is when LCOE = LACE for the next most valuable project, and we call this the equilibrium point. If demand grows between years, the most efficient plants will be dispatched more due to the merit order. Thus, the LACE for these plants will rise with demand, and new capacity of this technology will be added to the market until the equilibrium point is reached again.

Before considering capacity additions, the model will consider removing plants from technologies that are not economically viable to keep online. The observation that LACE < LCOE in a given year does not necessarily mean that plants will be removed, because plant managers may keep operating while they recover fixed and variable O&M costs: FOM + VOM < LACE. Moreover for any older technology, LACE < LCOE simply means that no more capacity will be added. However, when LACE < FOM, managers would gain more benefit (i.e., lose less) by shutting down the unit rather than operating it. During this removal process, for each eliminated megawatt of capacity, the remaining power plants will increase production to satisfy the demand, increasing their capacity factors and reducing their LCOE, until the point where the LACE for the lowest net value technology equals its fixed cost. Capacity expansion and contraction occurs in steps that represent an average power plant capacity in a technology group. Like the expansions decision, retirement decisions are myopic and only look at the current year’s LACE and FOM. Screening Curve Method8 The Screening Curve Method (SCM) uses annual load shape information together with the costs of competing power plant technologies, such as annualized capital costs and variable fuel costs, to find a least-cost generation mix solution. The actual construction dates of units in that portfolio as well as

8 For more details, see Zhang, Baldick, & Deetjan (2015) and Zhang & Baldick (2016). An executable version of the model is available at http://users.ece.utexas.edu/~baldick/screening_curve_method_tool/scm.html

17 those inevitable high price periods is critical to the VOM < LACE. Moreover for any older technology, model system behavior. To mimic the likely price LACE < LCOE simply means that no more capacity behavior in ERCOT, we used price data between will be added. However, when LACE < FOM, 2011 and 2015, when on average there were 531 managers would gain more benefit (i.e., lose less) hours per year (about 6%) with prices above $50 in by shutting down the unit rather than operating it. ERCOT, and an average of 18 hours where prices During this removal process, for each eliminated were $0 (Figure 6 of Potomac Economics, 2016). megawatt of capacity, the remaining power plants will increase production to satisfy the demand, We used a linear interpolation between the increasing their capacity factors and reducing provided data points to produce our price their LCOE, until the point where the LACE for distribution in Figure 5. To apply these prices, the lowest net value technology equals its fixed each hour was ranked in terms of its “thermal cost. Capacity expansion and contraction occurs stress” (load provided by thermal technologies / in steps that represent an average power plant installed thermal capacity). Our belief was that capacity in a technology group. Like the expansions prices would be the highest when the thermal decision, retirement decisions are myopic and technologies were utilized the most, so the hour only look at the current year’s LACE and FOM. with the highest thermal stress was assigned the highest price, here $9,000/MWh, the next highest Screening Curve Method 8 hour was assigned $4,500/MWh, and so on for the top 6% of hours in the thermal stress distribution. The Screening Curve Method (SCM) uses annual For the remaining 94% of hours, the marginal price load shape information together with the costs of the highest dispatched technology was applied of competing power plant technologies, such as with the exception that the 18 hours with the lowest annualized capital costs and variable fuel costs, thermal stress were assigned prices of $0/MWh. to find a least-cost generation mix solution. The actual construction dates of units in that Every time a new megawatt of generation portfolio as well as retirements might occur capacity is added, the capacity factor for all other at any year between the present and 2030. So, technologies drops because the new project’s the calculated generation mix is a cost-based generation will displace other power plants, result independent of market behavior. assuming demand is constant and there are no retirements. Therefore, for every new megawatt The SCM does not specify generator sizes, and constructed, LACE for all technologies will drop each generator is assumed to be the same (100 until there are no more profitable projects left to MW each in this exercise). A 1 GW generator add. This point is when LCOE = LACE for the is assumed to be the same as 10 generators next most valuable project, and we call this the of 100 MW capacity each. The generators are equilibrium point. If demand grows between years, assumed to be operated in one of three possible the most efficient plants will be dispatched more states at any given time: 100% output, 30% output due to the merit order. Thus, the LACE for these (corresponding to the minimum output level for plants will rise with demand, and new capacity the generator), or off. No other intermediate output of this technology will be added to the market levels are considered. The 30% level is chosen to until the equilibrium point is reached again. approximately correspond to typical minimum capacity for daily cycling units. Another minimum Before considering capacity additions, the model capacity level could be assumed instead.9 will consider removing plants from technologies that are not economically viable to keep online. The observation that LACE < LCOE in a given year does 8 For more details, see Zhang, Baldick, & Deetjan (2015) and Zhang & not necessarily mean that plants will be removed, Baldick (2016). An executable version of the model is available at http:// because plant managers may keep operating while users.ece.utexas.edu/~baldick/screening_curve_method_tool/scm.html they recover fixed and variable O&M costs: FOM + 9 In PLEXOS and AURORAxmp, each thermal unit has a minimum stable level ranging from 20% to 55% (Appendix A).

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 8 The unit commitment and economic dispatch and 25 years for wind based on Ventyx historical problems are approximated based on the trade- data. If a power plant is due to retire before off between maintaining a generator at minimum year 2030, it will be removed from the existing output level and shutting it down. To balance the capacity and counted as retirement. The SCM fluctuating load, the SCM calculates the associated also considers economic retirement when retrofit costs of operating at minimum output level versus cost plus variable fuel cost (VFC) is higher than shutting down and chooses the cheaper option. building and running a new unit of any technology type. For details, see Zhang and Baldick (2016). Unlike AURORAxmp and PLEXOS, but similar to the Excel model, the SCM aggregates all Each new technology can be viewed as a function the generators in different ERCOT zones to of load levels as illustrated in Figure 6, where form a single node. No transmission lines or three technologies are plotted (New Coal, New congestion are considered. The new generation CC, and New CT). The vertical axis represents technologies are assumed to be available in all the total cost, which includes VFC, VFC at zones; there is only one set of characteristics for minimum output level of 30%, Annualized each potential new unit. Also, the SCM assumes Capital Cost, Start-up Cost (SC), FOM and one load growth profile rather than regional VOM (see Zhang et al., 2015 for more details). profiles, and scales up ERCOT load to 81.2 GW peak in 2030, the same as other models. The lower load level corresponds to higher operating hours (equivalently higher capacity Unlike the other models, wind and solar resources factor) for a generating unit, which then are not dispatched. Using the same wind and corresponds to higher total cost (VFC times longer solar profiles as other models, the SCM treats operating hours). The three competing technologies them as negative load to calculate net load. The have different total cost curves because of different which includes VFC, VFC at minimum output level of 30%, Annualized Capital Cost, Start-up Cost (SC), portfolio of chosen generators is economically capital costs and VFCs. Since coal units have higher FOM and VOM (see Zhang et al., 2015 for more details). adapted to this net load. The 2030 wind and capital cost but lower VFC, it is more economical to solar capacity are the same as theThe LT lowerexpansion load level correspondsrun them tofor higher longer operating operating hours hours, (equivalently which occurs higher capacity factor) for a results from AURORAxmp (see resultsgenerating section unit, which thenat lower corresponds load levels. to higherAccordingly, total cost the (VFC total times cost longerof operating hours). The below). No hydro, demand response,three nor competing any technologiescoal units have is differentthe least totalexpensive cost curves among because the three of different capital costs and types of storage are modeled in theVFC SCM.s. Since Existing coal units havecandidates higher capital at lower cost load but levels. lower TheVFC, itSCM is more chooses economical to run them for thermal technologies are combinedlonger into operating six hours,the which least-cost occurs atsegments lower load from levels. the threeAccordingly, curves the total cost of coal units is the categories for this study; more categoriesleast expensive can be among theand three forms candidates a least-cost at curve.lower loadThe levels. crossing The pointsSCM chooses the least-cost used, but the result will not differsegments significantly. from the threewhere curves two and curves forms cross a least separate-cost curve. the generation The crossing points where two curves cross separate the generationsystem system into three into regions, three regions, each region each region balanced balanced by one technology. If the The long-term average forced outage rates (FORs) by one technology. If the whole system were to were obtained for different generationwhole systemtechnologies. were to be built from the ground up, the cost-minimizing portfolio would be about 41 GW FIGURE 6 For example, if the forced outageof rate new of coal a coal-fired, 24 GW of new CC, and 30 GW of new CT. Screening Curve Method for New Technologies generator is 10% and its capacity is 100 MW, it is Figure 6 – Screening Curve Method for New Technologies assumed that it can only reliably supply a 90-MW load. The operating and start-up costs are reflected by a 90-MW generation slice, while the fixed cost of 100 MW is accounted for in expansion costs.

Unlike AURORAxmp and the Excel model, which build or retire based on the economic value of individual units, the SCM assumes operational lives for each of the technologies: 65 years for coal power plants, 55 years for combined-cycle gas turbine (CC as shorthand for CCGT) and simple-cycle combustion turbines (CT), 60 years for nuclear,

Figure 7 – Screening Curve Method Considering Existing Technologies

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 9

The cost curves of the existing capacity can be calculated in a similar way except that their fixed costs are already sunk. Based on the cost curve shapes, the existing capacity blocks are then fitted into the screening curves of the new technologies. The objective of positioning the existing capacity is to minimize the overall system cost, the shaded area in Figure 7. Some of the potential new technology cost segments are replaced by the blocks of the existing technologies. Now, the least-cost curve (upper

19 which includes VFC, VFC at minimum output level of 30%, Annualized Capital Cost, Start-up Cost (SC), FOM and VOM (see Zhang et al., 2015 for more details).

The lower load level corresponds to higher operating hours (equivalently higher capacity factor) for a generating unit, which then corresponds to higher total cost (VFC times longer operating hours). The three competing technologies have different total cost curves because of different capital costs and VFCs. Since coal units have higher capital cost but lower VFC, it is more economical to run them for longer operating hours, which occurs at lower load levels. Accordingly, the total cost of coal units is the least expensive among the three candidates at lower load levels. The SCM chooses the least-cost segments from the three curves and forms a least-cost curve. The crossing points where two curves cross separate the generation system into three regions, each region balanced by one technology. If the whole system were to be built from the ground up, the cost-minimizing portfolio would be about 41 GW of new coal, 24 GW of new CC, and 30 GW of new CT.

Figure 6 – Screening Curve Method for New Technologies

FIGURE 7 the potential new CC is completely replaced by the FigureScreening 7 – Screening Curve Method Curve Method Considering Considering Existing ExistingTechnologies Technologies existing capacity, which means that there will be no new CC built for this particular set of assumptions. Summary of Key Differences across the Models

There are some key differences across the modeling approaches (Table 1). Two commercial software programs (AURORAxmp and PLEXOS) represent every single generating unit in a system and their operational characteristics. In contrast, existing generation units are grouped into 10 technology sets in the Excel model and 6 thermal technologies for net load in SCM. New technology groups are also treated differently: eight in AURORAxmp, six in the Excel model, and six The cost curves of the existing capacity can be calculated in a similar way except that their fixed costs be built from the ground up, the cost-minimizing thermal technologies in the SCM. Eight zones are are already sunk. Based on theportfolio cost curve would shapes, be about the existing41 GW capacityof new coal, blocks are thenmodeled fitted into in theAURORAxmp and PLEXOS versus a screening curves of the new24 technologies. GW of new The CC, objective and 30 GW of positioning of new CT. the existing capacitysingle ERCOTis to zone in the Excel model and SCM. minimize the overall system cost, the shaded area in Figure 7. Some of the potential new technology cost segments are replaced Theby the cost blocks curves of the of theexisting existing technologies. capacity canNow, be the least-Thecost curvefundamental (upper economic concepts used in the calculated in a similar way except that their fixed Excel model and the dispatch software programs costs are already sunk. Based on the cost curve are similar but different19 in potentially critical shapes, the existing capacity blocks are then fitted ways. For the long-term capacity expansion, the into the screening curves of the new technologies. algorithm of AURORAxmp (MIP with value The objective of positioning the existing maximization) searches for the mix of resources capacity is to minimize the overall system cost, the that would maximize the overall profit, building shaded area in Figure 7. Some of the potential new new or retiring existing units simultaneously based technology cost segments are replaced by the blocks on NPV over the planning horizon. In contrast, of the existing technologies. Now, the least-cost the Excel model retires a certain capacity of a curve (upper bound of the shaded area) is a piece- technology for which LACE is less than the FOM, wise curve that consists of both new technologies and then decides to build new generation to meet and existing technologies. Note that, in this case, demand as long as LACE is greater than LCOE in

TABLE 1 Key Differences across Four Models

Technologies New Builds Retirements Renewables Zones* AURORAxmp Individual units NPV over planning NPV over planning Dispatched via wind & solar 8 horizon horizon shapes PLEXOS Individual units N/A N/A Dispatched via wind & solar 8 shapes Excel All units grouped into LACE > LCOE per LACE < FOM per year Dispatched via wind & solar 1 10 existing and 6 new year shapes technology groups SCM All units grouped into 6 Least-Cost VFC Deducted from load to obtain 1 existing and 6 new ther- net load mal technology groups * There are different assumptions for energy and peak load growth, some generation characteristics across the zones. Transmission capacity between the zones impacts congestions and zonal prices, which can be important for new build or retirement decisions.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 10 TABLE 2 Resource Capacity by Fuel Type and Primary Mover

Fuel Type Net Capacity 2015 (MW) Primary Mover Biogas 93.5 Internal Combustion Engine Biomass 150.0 Steam Turbine Coal, lignite 7,142.0 Steam Turbine Coal, subbituminous 12,637.0 Steam Turbine Natural Gas 51,551.6 Total 34,629.2 Combined-Cycle Gas Turbine 11,970.0 Steam Turbine 4,724.8 Open-Cycle Gas Turbine 227.6 Internal Combustion Engine Solar 287.7 Photovoltaic Water 555.1 Hydraulic Turbine Wind 15,887.1 Wind Turbine Uranium 5,133.0 Steam Turbine TOTAL* 93,437.0

* Note that this number represents installed capacity and is significantly larger than operational capacity reported in ERCOT (2015c). There are two main reasons. First, roughly 8,700 MW of combined heat and power (CHP) capacity is included (see detailed description of these units below). Second, we report nameplate capacity for wind; ERCOT assumes 14% peak average capacity contribution for non-coastal wind and 58% for coastal wind. We calculate annual hourly wind profiles for use in dispatch models (see below).

a given year. Both the Excel model and the SCM cost generation portfolio for a target year using are designed to minimize total cost. Because the six existing and six new thermal technologies. Excel model considers ERCOT as a single zone, The renewables are not dispatched but are a 35% “availability factor” for natural gas was deducted from load to obtain net load. imposed to mimic the transmission constraints in the actual system. In other words, natural gas COMMON INPUTS plants cannot run more than 35% of the time. In the following sections, we describe parameters The SCM is a different approach in that, in its and assumptions reconciled among all four simplest form, does not separate commitment models, as needed, for 2015. This is the common and dispatch, but rather incorporates the cost departure point for the long-term simulations of dispatch into the annualized cost model. In and for the 8,760 hourly dispatch runs in 2030. more advanced versions, SCM considers daily Resources commitment based on the potential “off-line” time duration of each generator. In the version The net capacity in ERCOT through the of SCM used for this study, if the potential off- end of 2015 is summarized in Table 2. The line time is less than 8 hours, then the generator operational parameters for each individual is maintained at minimum production during unit included in this database were gathered these hours; if the duration is more than 8 hours, and reconciled from various sources.11 the unit is shut down.10 SCM seeks the least-

11 The following sources were referenced: (EIA 2015), (EIA 2016a), (EPA 2016), (ERCOT 2015c), (FERC 2016), (ICF 2016), (PUCT 2015), (TCEQ 10 More sophisticated dynamic thresholds can be incorporated, as 2015), (TCEQ 2016), (TCPA 2016). The list of units and key operational described in Zhang et al. (2015). parameters are provided in Appendix A.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 11 Fuels the updated 2014 dataset since a 2015 dataset was not available at the time of conducting the analysis. The following thermal generator fuels are included in our baseline model (Table 3). The ERCOT separates wind generators into three EIA Form 923 fuel code is listed in parentheses. categories: inland, coastal and offshore. However, For the model runs, we used price projections we wanted to strike a balance between reduced for natural gas, lignite and subbituminous coal modeling complexity and capturing more (Figure 26 and Figure 27); the numbers in Table localized variance for the implementation of 3 are for the purposes of a quick comparison. the AURORAxmp and PLEXOS models. To Other fuels do not have a significant market do this, we took the capacity-weighted average share (biogas and biomass) or the price does of all wind generators in each county to create not change significantly over time to impact county-level output profiles. These were then capacity factor of the plants (uranium). normalized between 0 and 1 and used as rating factors (% of maximum output at each hour). TABLE 3 Fuels Included in Modeling Some of the counties did not contain any wind sites and thus lacked any output profiles. In this Baseline Price case, we created composite profiles based on the Category Fuel [$2015/MMBtu] output profiles of adjacent counties (Figure 8). We Biogas Landfill gas (LFG) - used a minimum of two different counties for each Wood and wood waste solids composite county. Where no adjacent counties were Biomass 0.05 (WDS) available, we continued the search radially outward. Coal Lignite coal (LIG) 2.68 Only three composite counties needed two hops, Coal Subbituminous coal (SUB) 2.07 and just one county needed more than two hops. Gas Natural gas (NG) 2.63 FIGURE 8 Uranium Uranium (NUC) 0.51 Composite Wind Counties. Composite counties Some generators may use secondary fuels in orange; source counties in yellow. for restart, backup or market reasons. These fuels, including some biogas, agricultural byproducts, black liquor, petroleum coke, distillate fuel oil, and jet fuel, typically represent a small fraction of overall fuel consumption in ERCOT, and they are not likely to increase their market share in the future. As such, we have excluded them for simplicity.

Wind

Wind generator outputs are based on an hourly wind plant dataset produced for ERCOT by AWS Truepower (ERCOT, 2015b; AWS, 2012). The dataset consists of hourly wind generator output profiles for existing and hypothetical wind sites throughout Texas from 1997 through 2014. These profiles were generated using the Weather Research and Forecasting Model, a mesoscale numerical Finally, we composed three representative annual weather prediction model, alongFinally, with wecomposite compose d threehourly representative wind profiles annual for hourlyERCOT wind load profiles zones for ERCOT load zones (Figure 9) power curves for different turbinebased classes. on county We -usedlevel data,(Figure including 9) based ERCOT on Northcounty-level (Summary data, statistics including are provided in Table 4. There is significant variability of capacity factor across the year and some variability across the three load zones. For example, there are hours where there is no wind generation in each of the load zones and there are hours where wind generation can be above 90% of installed capacity, with an annual average of about The Full Cost of Electricity (FCe-) 40% in the North and West loadModel zones Documentation and 36% and Results in the for SouthERCOT Scenarios, load zone. July 2017 | 12

However, the wind output distributions are not normal nor identical across the three zones. The distribution for the North zone is U-shaped: there are more instances of very low and very high capacity factors than the average values. The distributions for the South and West zones are right-skewed: most of the observations fell below the average capacity factor. On average, capacity factors are higher in June (50% to 56%) than in September (23% to 30%). There is also variability across counties; a version of Table 4 by county is provided in Appendix C: Wind Rating Factors by Aggregate County (%).

Figure 10), ERCOT South (Figure 11), and ERCOT West (Figure 12), as inputs for the subsequent long- term modeling.13

Figure 9 – ERCOT Load Zones

13 The allocation of counties into these load zones is provided in the generator list of Appendix C: Wind Rating Factors by Aggregate County (%). The data for a handful of Houston counties are also reported although we do not discuss those in this section. There are a limited number of projects in the Houston zone.

23

ERCOT North (Figure 10), ERCOT South (Figure 56%) than in September (23% to 30%).13 There 11), and ERCOT West (Figure 12), as inputs is also variability across counties; a version of for the subsequent long-term modeling.12 Table 4 by county is provided in Appendix C.

FIGURE 9 FIGURE 10 ERCOT Load Zones 2014 Hourly Wind Profile for ERCOT North Load Zone

100% ERCOT North 90% 100%80% ERCOT North 90%70% 100% 80%60% ERCOT North 90% 70%50% 80% 60%40% 70% 50%30% 60% Rating Factor Factor [% of Max] Rating 40%20% 50% 30%10% 40%0%

Rating Factor Factor [% of Max] Rating 20% 30% 10%

Rating Factor Factor [% of Max] Rating 20% 0% 10% 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 0%

20140101 20140114 Figure20140128 20140210 1120140224 – 201420140309 20140323 Hourly20140405 20140419 20140502 Wind20140516 P20140529 rofile20140612 20140626 for20140709 ERCOT20140723 20140805 South20140819 20140901 Load20140915 20140928 Zone20141012 20141025 20141108 20141122 20141205 20141219 FIGURE 11

100% 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 2014 HourlyFigure Wind 11 Profile – 2014 Hourlyfor ERCOT Wind ProfileSouth South for Load ERCOT Zone South Load Zone 90% Figure 11 – 2014 Hourly Wind Profile for ERCOT South Load Zone 100%80% ERCOT South 90%70% 100% ERCOT South 80%60% 90% Source: ERCOT Source: ERCOT 70%50% 80% 60%40% 70% 50%30% Summary statistics are provided in Table 4. There 60% Summary statistics are provided in Table 4. There is significant variability of capacity factor Factor [% of Max] Rating 40%20% across the is significant variability of capacity factor across 10%50% year and some variability across the three load zones. For example, there are hours where30% there is no 40%0% Rating Factor Factor [% of Max] Rating 20% the year and some variability across the three 30% wind generation in each of the load zones and there are hours where wind generation can10% be above 90% load zones. For example, there are hours where Factor [% of Max] Rating 20% 0% of installed capacity, with an annual average of about 40% in the North and West load zones10% and 36% in there is no wind generation in each of the load 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 the South load zone. 0% zones and there are hours where wind generation

20140101 20140114 Figure20140128 20140210 1220140224 – 201420140309 20140323 Hourly20140405 20140419 20140502 Wind20140516 20140529 Profile20140612 20140626 for20140709 ERCOT20140723 20140805 West20140819 20140901 Load20140915 20140928 Zone20141012 20141025 20141108 20141122 20141205 20141219 can be above 90% of installed capacity, with an

However, the wind output distributions are not normal nor identical across the three zones. The20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 annual average of about 40% in the North and FIGURE100% 12Figure 12 – 2014 HourlyERCOT Wind Profile West for ERCOT West Load Zone distribution for the North zone is U-shaped: there are more instances of very low and very90% high capacity West load zones and 36% in the South load zone. 2014 Hourly FigureWind 12 Profile – 2014 Hourlyfor ERCOT Wind ProfileWest forLoad ERCOT Zone West Load Zone factors than the average values. The distributions for the South and West zones are right100%-80%skewed: most ERCOT West 90%70% 100% ERCOT West of the observations fell below theHowever, average the capacity wind output factor. distributionsOn average, capacityare not factors are80%60% higher in 90% 14 70%50% June (50% to 56%) than in Septembernormal (23% nor identicalto 30%). across There the is also three variability zones. The across counties;80% a version 60%40% distribution for the North zone is U-shaped: 70% of Table 4 by county is provided in Appendix C: Wind Rating Factors by Aggregate County 50%30%(%).

Rating Factor Factor [% of Max] Rating 60% there are more instances of very low and very 40%20% 50% 30%10% high capacity factors than the average values. 40% Rating Factor Factor [% of Max] Rating 20%0% 30% The distributions for the South and West zones 10% Rating Factor Factor [% of Max] Rating 20% 0% are right-skewed: most of the observations fell 10% 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 below the average capacity factor. On average, 0% Figure 10 – 2014 Hourly Wind Profile for ERCOT North Load Zone capacity factors are higher in June (50% to 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 25 20140101 20140114 20140128 20140210 20140224 20140309 20140323 20140405 20140419 20140502 20140516 20140529 20140612 20140626 20140709 20140723 20140805 20140819 20140901 20140915 20140928 20141012 20141025 20141108 20141122 20141205 20141219 25 13 Note that these observations are based on the 2014 data used in this

analysis. Historically, the highest wind generation in Texas has occurred 25 12 The allocation of counties into these load zones is provided in the in spring or fall months, and the lowest generation in the summer generator list of Appendix C. The data for a handful of Houston counties months. However, these periods vary over the years along with weather are also reported although we do not discuss those in this section. There patterns. Note that the highest-energy day and the lowest-energy day are a limited number of projects in the Houston zone. are more consistent with historical expectations.

14 Note that these observations are based on the 2014 data used in this analysis. Historically, the highest wind generation in Texas has occurred in spring or fall months, and the lowest generation in the summer months. The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 13 However, these periods vary over the years along with weather patterns. Note that the highest-energy day and the | lowest-energy day are more consistent with historical expectations.

24

TABLE 4 Wind Rating Factors by Load Zone (%)

Annual Table 4 – Wind JuneRating Factors by Load Zone (%)September Load Zone Min Mean Max AnnualMin Mean Max June Min MeanSeptemberMax Load Zone Min Mean Max Min Mean Max Min Mean Max North 0.0 North40.3 93.2 0.0 40.3 0.1 93.255.2 0.192.5 55.2 92.50.0 22.90.0 22.989.3 89.3 South 0.0 South35.7 91.9 0.0 35.7 0.7 91.950.1 0.790.4 50.1 90.40.1 26.70.1 26.782.3 82.3 West 0.0 40.7 91.8 1.1 55.8 90.0 0.7 29.9 87.2 West 0.0 40.7 91.8 1.1 55.8 90.0 0.7 29.9 87.2

The highest wind-energy day (aggregate ERCOT average) in 2014 occurs on March 26, a shoulder month The highest wind-energywith low total day system(aggregate demand (Figure 13The). Although lowest-energy wind generation day in 2014 is fairly occurs high on and May stable throughout ERCOT average)the in day2014 in occursthe West on andMarch North load zones29, (70% which-90% might rating be factor), considered it starts the low beginning (about 43%) of in the South 26, a shoulder monthload zone with and low peaks total atsystem about 81% ratingsummer factor in in mid Texas-day (Figurebefore starting14). Again, to decline. the South load demand (Figure 13). Although wind generation zone has a somewhat different profile than the other is fairly high andThe stable lowest throughout-energy day the in day 2014 occurs onzones, May 29, but which wind mightgeneration be considered is very low the in beginning all three of summer in the West and inNorth Texas load (Figure zones 14 ).(70%-90% Again, the South loadload zone zones. has aThe somewhat highest different rating factor profile occurs than inthe the other zones, rating factor), it butstarts wind low generation (about 43%) is very in the low in all threeSouth load load zones. zone The at highestabout 17% rating in factorthe early occurs evening. in the South load South load zonezone and peaksat about at about17% in 81% the earlyrating evening. The North North load load zone zone peaks peaks at atabout about 5% 5% and and the the West load zone factor in mid-daypeaks before at aboutstarting 10% to indecline. the evening. West load zone peaks at about 10% in the evening.

Figure 13 – Wind Energy Production on Highest-Energy Day FIGURE 13 March 26th Wind Energy Production 100% on Highest-Energy Day 80%

60%

40%

Rating Factor Factor [% of Max] Rating 20%

0% 0:00 6:00 12:00 18:00 0:00 North LZ South LZ West LZ

Figure 14 – Wind Energy Production on Lowest-Energy Day FIGURE 14 May 29th Wind Energy Production 18% on Lowest-Energy Day 16% 14% 12% 10% 8% 6%

Rating Factor Factor [% of Max] Rating 4% 2% 0% 0:00 6:00 12:00 18:00 0:00 North LZ South LZ West LZ

26 Solar The Full Cost of ElectricityThe solar (FCe-) generation profile used for this studyModel was Documentation created andby Results approximating for ERCOT Scenarios, the locationJuly 2017 | and 14 distribution of future solar resources and combining their unique solar generation profiles (Table 5). The ERCOT “Generator Interconnection Status Report” for February 2015 (ERCOT 2015a) was used to find the county location of proposed solar projects.

Table 5 – The Distribution of Solar Resources in ERCOT Location % of Total Capacity Marfa 55.38 Midland 22.12 Laredo 7.04 Amarillo 5.49 Abilene 4.75 Del Rio 3.38 Dallas 1.83

A unique generation curve was calculated for each location using the PVWatts calculator published by the National Renewable Energy Laboratory (NREL 2016). This calculator uses information about the solar array along with typical meteorological year (TMY) data to calculate an array’s hourly solar output over one year. The arrays in this study were modeled as one-axis tracking arrays with 96% efficient inverters and a 1.1 DC-to-AC size ratio.

Figure 15 – Typical Winter Day Solar PV Production by Region

27

Solar TABLE 5 Distribution of Solar Resources in ERCOT The solar generation profile used for this study was created by approximating the location Location % of Total Capacity and distribution of future solar resources Marfa 55.38 and combining their unique solar generation Midland 22.12 profiles (Table 5). The ERCOT “Generator Laredo 7.04 Interconnection Status Report” for February 2015 (ERCOT 2015a) was used to find the Amarillo 5.49 county location of proposed solar projects. Abilene 4.75 100% Del Rio January 18th 3.38 A unique generation curve was calculated for each 90% Dallas 1.83 location using the PVWatts calculator published80% by the National Renewable Energy Laboratory70% (NREL and summer day, respectively. In winter, more 2016). This calculator uses information 60%about the frequent cloudy weather reduces the rating solar array along with typical meteorological50% year factor and increases the volatility of most of (TMY) data to calculate an array’s hourly40% solar the solar arrays. Table 6 shows the statistical output over one year. The arrays in this 30%study were distribution of rating factors for each array. modeled as one-axis tracking arrays with20% 96% efficient inverters and a 1.1 DC-to-AC Factor [% of Max] Rating 10%size ratio. By adding the generation curves together and 0% weighting them based on the interconnection Figure 15 and Figure 16 show the capacity 0:00 distribution6:00 (Table 5), 12:00a conglomerate solar18:00 Marfa Midland Laredo Amarillo factors for each of the arrays on a typical winter generation curve was created and used in Abilene Del Rio Dallas FIGURE 15 100% January 18th Typical Winter Day Solar PV Figure 15 and Figure90% 16 show the capacity factors for each of the arrays on a typical winter and summer Production by Region day, respectively.80% In winter, more frequent cloudy weather reduces the rating factor and increases the volatility of most 70%of the solar arrays. Table 6 shows the statistical distribution of rating factors for each array. 60% 50% By adding the generation40% curves together and weighting them based on the interconnection distribution (Table 5), a conglomerate30% solar generation curve was created and used in the models. This conglomerate curve20% can be scaled up to any installed capacity to produce a representative ERCOT solar generation curve. Factor [% of Max] Rating 10% Figure 17 shows the typical winter solar generation curve. Capacity factors tend to be 0% lower in the winter than0:00 in the summer6:00 due to the lower12:00 angle of the sun18:00 and the shorter amount of daylight. Marfa Midland Laredo Amarillo Abilene Del Rio Dallas Figure 16 – Typical Summer Day Solar PV Production by Region FIGURE 16 100% June 9th Typical Summer Day Solar Figure 15 and Figure 16 show the capacity factors for each of the arrays on a typical winter and summer 90% PV Production by Region day, respectively. In winter, more frequent cloudy weather reduces the rating factor and increases the 80% volatility of most 70%of the solar arrays. Table 6 shows the statistical distribution of rating factors for each array. 60% 50% By adding the generation curves together and weighting them based on the interconnection distribution 40% (Table 5), a conglomerate30% solar generation curve was created and used in the models. This conglomerate curve20% can be scaled up to any installed capacity to produce a representative ERCOT solar

generation curve. Factor [% of Max] Rating 10% Figure 17 shows the typical winter solar generation curve. Capacity factors tend to be lower in the winter0% than in the summer due to the lower angle of the sun and the shorter amount of 0:00 6:00 12:00 18:00 daylight. Marfa Midland Laredo Amarillo

Figure 16Abilene – Typical SummerDel Rio Day Solar DallasPV Production by Region

100% June 9th Table 6 – Solar Rating Factors by Region (%) The Full Cost of Electricity (FCe-) 90% Model Documentation and Results for ERCOT Scenarios, July 2017 | 15 Annual Winter Summer 80% Min Mean Max Min Mean Max Min Mean Max Midland 0.0 70% 24.0 91.0 0.0 20.0 91.0 0.0 25.0 76.0 Marfa 0.0 60% 24.0 83.0 0.0 17.0 77.0 0.0 30.0 81.0 50% 40% 28 30% 20%

Rating Factor Factor [% of Max] Rating 10% 0% 0:00 6:00 12:00 18:00 Marfa Midland Laredo Amarillo Abilene Del Rio Dallas

Table 6 – Solar Rating Factors by Region (%) Annual Winter Summer Min Mean Max Min Mean Max Min Mean Max Midland 0.0 24.0 91.0 0.0 20.0 91.0 0.0 25.0 76.0 Marfa 0.0 24.0 83.0 0.0 17.0 77.0 0.0 30.0 81.0

28

TABLE 6 Solar Rating Factors by Region (%)

Annual Winter Summer Min Mean Max Min Mean Max Min Mean Max Midland 0.0 24.0 91.0 0.0 20.0 91.0 0.0 25.0 76.0 Marfa 0.0 24.0 83.0 0.0 17.0 77.0 0.0 30.0 81.0 Amarillo 0.0 22.0 84.0 0.0 15.0 71.0 0.0 28.0 79.0 Dallas 0.0 20.0 79.0 0.0 13.0 71.0 0.0 26.0 77.0 Abilene 0.0 22.0 82.0 0.0 15.0 70.0 0.0 27.0 75.0 Del Rio 0.0 18.0 77.0 0.0 13.0 70.0 0.0 24.0 74.0 Laredo 0.0 20.0 81.0 0.0 13.0 71.0 0.0 25.0 76.0 Amarillo 0.0 22.0 84.0 0.0 15.0 71.0 0.0 28.0 79.0 the models. This conglomerateDallas curve0.0 can be20 .0 79.0inverter losses,0.0 the13 projection.0 71 .0of the sun 0.0onto the26 .0 77.0 scaled up to any installedAbilene capacity 0.0to produce22.0 a 82.00-degree horizontal0.0 15 panels.0 used70.0 by 1-axis 0.0 tracking 27 .0 75.0 representative ERCOTDel solar Rio generation0.0 curve.18.0 77.0arrays, and0.0 other inefficiencies13.0 70.0 in the system.0.0 24.0 74.0 Laredo 0.0 20.0 81.0 0.0 13.0 71.0 0.0 25.0 76.0 Figure 17 shows the typical winter solar generation curve. Capacity factors tend to be lower inFigure the 17 – TypicalCombined Winter Day Conglomerate Heat and Power Solar PV Plants Production by Region winter than in the summerAmarillo due to the0.0 lower22 angle.0 84.0 0.0 15.0 71.0 0.0 28.0 79.0 Dallas 0.0 20.0 79.0 0.0 13.0 71.0 0.0 26.0 77.0 1.0 Industrial combined heat and power plants of the sun and the shorterAbilene amount 0.0of daylight.22.0 82.0 0.0January 15 .018th 70.0 0.0 27.0 75.0 Del Rio 0.0 18.0 77.0(CHP), also0.0 known13 as.0 cogeneration70.0 plants,0.0 are 24.0 74.0 Figure 18 shows the typicalLaredo summer0.0 solar 0.8 20.0 81.0designed to0.0 supply 13 heat.0 to industrial71.0 processes0.0 25.0 76.0 generation curve. The peak is less than 1.0 due to and to allow excess heat to produce electricity. 0.6 Figure 17 – Typical Winter Day Conglomerate Solar PV Production by Region FIGURE 17 1.0 Typical Winter Day Conglomerate 0.4 January 18th Solar PV Production by Region 0.8 0.2

0.6 MW Generated per MW per Installed Generated MW 0.0 0:00 6:00 12:00 18:00 0.4 Marfa Midland Laredo Amarillo Abilene Del Rio Dallas

0.2 Figure 18 shows the typical summer solar generation curve. The peak is less than 1.0 due to inverter losses, the projectionMW per Installed Generated MW 0.0 of the sun onto the 0-degree horizontal panels used by 1-axis tracking arrays, and other inefficiencies in0:00 the system. 6:00 12:00 18:00 Marfa Midland Laredo Amarillo Abilene Del Rio Dallas Figure 18 – Typical Summer Day Conglomerate Solar Production by Region FIGURE 18 1.0 Typical Summer Day ConglomerateFigure 18 shows the typical summer solar generationJune curve.9th The peak is less than 1.0 due to inverter Solar Production by Regionlosses, the projection of the sun onto the 0-degree horizontal panels used by 1-axis tracking arrays, and other inefficiencies0.8 in the system.

0.6 Figure 18 – Typical Summer Day Conglomerate Solar Production by Region

1.0 0.4 June 9th

0.8 0.2

0.6 MW Generated per MW per Installed Generated MW 0.0 0:00 6:00 12:00 18:00 0.4 Marfa Midland Laredo Amarillo Abilene Del Rio Dallas

0.2 The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 16 Combined Heat and Power Plants |

Industrial combinedMW per Installed Generated MW 0.0 heat and power plants (CHP), also known as cogeneration plants, are designed to supply heat to industrial0:00 processes and6:00 to allow excess12:00 heat to produce 18:00electricity. The CHP plants in ERCOT can choose whenMarfa to useMidland power themselvesLaredo Amarilloand when Abileneto sell to theDel Riomarket;Dallas in other words, they

are self-dispatched. In some cases, the amount of capacity available for selling to the market may only Combinedbe a fraction Heat of andthe netPower power Plants capacity of a CHP plant. Industrial combined heat and power plants (CHP), also known as cogeneration plants, are designed to 29 supply heat to industrial processes and to allow excess heat to produce electricity. The CHP plants in

ERCOT can choose when to use power themselves and when to sell to the market; in other words, they are self-dispatched. In some cases, the amount of capacity available for selling to the market may only be a fraction of the net power capacity of a CHP plant.

29

The CHP plants in ERCOT can choose when to average of EIA Form 860 summer and winter use power themselves and when to sell to the net capacities. Finally, the rated capacity was market; in other words, they are self-dispatched. multiplied by the average yearly percent electricity In some cases, the amount of capacity available generation to obtain the dispatchable capacity for selling to the market may only be a fraction for modeling purposes (Table 7). At close to of the net power capacity of a CHP plant. 9,000 MW dispatchable capacity, CHP represents a significant portion of the ERCOT market. Following a method developed previously (Garrison, 2014), the yearly percent electricity Switchable Plants generation was calculated for each CHP plant based on EIA Form 923 data for years 2010– A switchable generator is one located adjacent to 14. Plants with zero to minimal electricity two synchronous grids that has interconnection generation over this period were ignored. Next, agreements with two different balancing the average yearly percent electricity generation authorities. As of 2015, there were four power was calculated for these five years. The rated plants in ERCOT that could be switched to other capacity was then calculated by taking the synchronous grids: Frontera Generation (CFE/

TABLE 7 Capacities for Combined Heat and Power Plants in ERCOT

2014 Mean Rated 2010-2014 Mean 2015 Dispatchable Capacity [MW] Generation [%] Capacity [MW] Bayou Cogen Plant 309.0 53.6% 165.6 Baytown Energy Center 807.5 98.4% 794.8 Baytown Energy Center Chiller Upgrade (2016) 270.0 98.4% 265.7 BP Chemicals Green Lake Plant 38.8 20.7% 8.0 C R Wing Cogen Plant 212.0 98.0% 207.7 Channel Energy Center 796.0 86.3% 686.6 Channelview Cogen Plant 865.0 99.5% 861.0 Clear Lake Cogeneration Ltd 384.9 100.0% 384.9 Corpus Christi Energy Center 468.0 98.4% 460.5 Deer Park Energy Center 1 1,152.0 100.0% 1,152.0 Equistar Corpus Christi 37.0 32.5% 12.0 ExxonMobil Baytown Refinery 157.5 2.7% 4.3 Freeport Energy Center 239.1 26.5% 63.4 Green Power 2 836.0 34.4% 287.2 Gregory Power Facility 388.5 100.0% 388.5 Houston Chemical Complex Battleground 281.5 18.7% 52.6 Ingleside Cogeneration 484.0 98.8% 478.4 Optim Energy Altura Cogen 580.3 82.5% 478.8 Oyster Creek Unit VIII 404.5 100.0% 404.5 Pasadena Cogeneration 762.5 97.5% 743.6 Sweeny Cogen Facility 470.0 91.6% 430.7 Texas City 1 467.5 98.1% 458.5 Texas Gulf Sulphur (New Gulf) 78.3 75.2% 58.9 Victoria Texas Plant 81.5 10.9% 8.9 Wichita Falls Cogeneration Plant 78.0 100.0% 78.0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 17 Mexico), Tenaska Frontier (MISO), Tenaska Hydroelectric capacity is relatively small in Gateway (SPP) and Tenaska Kiamichi (SPP). ERCOT (Table 2). Therefore, these assumptions Frontera Generation has announced that it will are not likely to have a significant impact on connect solely to CFE starting in 2016, so it will long-term capacity expansion runs, but they be excluded from the ERCOT area in our models. are important to capture for 8,760 runs. Additionally, the Antelope Elk Energy Center is being built in phases and will be switchable TABLE 8 between ERCOT and SPP. Although these plants Monthly Maximum Capacity Factors for may technically be able to switch between grids Hydroelectric Generators in ERCOT given sufficient lead time, we assume that they are Month Max Capacity Factor [%] all available to ERCOT year-round. Shell has stated January 11.2 that it connects the Tenaska Frontier and Tenaska Gateway plants to ERCOT the vast majority of the February 20.8 time under its tolling agreements (FERC 2011). March 36.4 April 32.2 Mothballed Plants May 33.6 ERCOT allows plants to enter a temporary June 31.0 mothball status rather than permanently shutting July 21.6 down. In these cases, a generator can be called August 12.6 upon to “provide voltage support, stability September 8.4 or management of localized transmission October 8.8 constraints” (ERCOT 2016c) until a remedial November 10.4 transmission project can be completed or the reliability concerns are otherwise mitigated. December 13.6 ERCOT usually does this by entering into reliability must-run contracts with these plants. However, these contracts are meant to be stop- Energy Storage gap solutions, and plants that are mothballed There are two large-scale battery systems operating rarely become full-time generators again. in ERCOT: the Presidio NaS Battery and the Mothballed units, especially if they are coal-fired, Notrees Battery Facility. The Presidio facility is require at least several weeks of advance notice designed for reliability only. The 36-MW NoTrees to get ready. For simplicity, we assume that any Battery Facility is co-located with the Notrees unit mothballed through 2015 or announced Wind Farm, a 152.6 MW nameplate wind farm to be mothballed is permanently retired. We do in Winkler County. It is set up to charge from the not attempt to model a mothball status in LT output of the wind turbines only, but it can then capacity expansion (840 MW of coal capacity discharge back into the grid. This captive setup in the future). These units represent a small does not allow the batteries to be charged from the percentage of the installed capacity in ERCOT. grid. Battery capacity is insignificant relative to the size of the ERCOT market and is not modeled Water in the base scenarios discussed in this paper. Hydroelectric generators are modeled based on Market structure maximum monthly capacity factors from EPIS (Table 8). The models then choose when to ERCOT has a nodal market design. There are dispatch based on hourly conditions. Since the thousands of electrical buses (nodes). Locational capacity factor is calculated from energy produced, marginal prices (LMPs) are reported for each the total capacity factor sets the maximum of these nodes every 5 minutes, consistent with amount of energy produced each month. security-constrained economic dispatch (SCED). These LMPs are aggregated to calculate prices

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 18 at about 630 settlement points. These settlement to compensate for variability in wind and solar prices can be different between certain nodes, generation, especially when the shares of these sometimes significantly so. These differences are technologies reach certain threshold levels. often due to weather anomalies, unscheduled generation or transmission outages, or transmission ERCOT does not have a long-term capacity market. bottlenecks. In daily grid operations, these Instead, it has instituted an operating reserve price signals are extremely important for the demand curve (ORDC) as of June 2014 to provide system operator to maintain reliability while additional revenues to marginal resources during meeting load in all zones at least cost. tight market conditions. AURORAxmp captures this market function by providing a price adder to Nevertheless, most of the time, prices are roughly units that may be needed to meet operating reserve the same across the ERCOT grid. Persistently margin in a zone. However, given the short history significant price differences are recognized by of the ORDC in ERCOT, we do not have sufficient market participants as arbitrage opportunities and data to represent it confidently via a price adder they are mitigated by the appropriate investment in in the model. According to Potomac Economics generation, transmission and/or demand response. (2016), the ORDC Adder averaged $1.41/MWh As such, in a 15-year capacity expansion run, a in 2015, its first full year of implementation. The nodal model does not add much value compared average was raised owing to high value in August to a zonal model. Accordingly, we use the zonal 2015 (about $9) but, the ORDC adder is expected version of AURORAxmp for long-term capacity to fluctuate from year to year subject to market expansion analysis (Figure 1). The Excel model and conditions. As such, one year’s data is hardly the SCM treat ERCOT as a single zone (or node). sufficient to develop long-term assumptions. Also, ERCOT (2016a) indicates the ERCOT market The hourly dispatch models of AURORAxmp and currently has more than adequate reserves for the PLEXOS are designed to simulate the day-ahead next several years, which would likely suppress the energy market. The real-time energy market is not value of the ORDC adder as well as other ancillary modeled since it focuses on short-term reliability at services. Therefore, we did not impose a price adder timescales of seconds to minutes. Both models can to reflect the scarcity of reserve margin in the runs include four types of ancillary services for reserves: discussed in this paper but, it is straightforward Regulation Up, Regulation Down, Spin and Non- to incorporate as a sensitivity in future analysis. Spin. We have been unable to identify the resources that provided such services on a regular basis and AURORAxmp and PLEXOS also include Demand revenues generated from such services. On the basis Side Curtailment resources, which may be of reporting by the independent market monitor,14 dispatched when energy output from all other we estimate that the per-MWh value of these resources does not meet zonal loads. For the ancillary services has been only 4% of the energy purposes of this study, we priced Demand Side price between January 2011 and December 2015. Curtailment resources at $9,000/MWh, the same as the current energy price cap in ERCOT. Our goal Accordingly, we exclude ancillary services in is to make sure that the model will try to dispatch long-term optimization for the base case scenarios all available resources, as well as exhaust all new presented in this paper. This way, AURORAxmp build options that are economically feasible, before results are more readily comparable to results it decides to dispatch a Demand Side Curtailment from the Excel model and the SCM, which resource. According to ERCOT (2016a), standard do not include ancillary services. We do note, offer load management programs amount to 208 however, that these services can become more MW, and emergency response service to 1,507 MW. significant in the future to capture the value of There is also 1,153 MW of load resources providing storage technologies as well as balancing provided responsive reserve services. Although these capacities appear low, they can still have a significant

14 http://www.potomaceconomics.com/index.php/markets_monitored/ impact on peak prices. We will incorporate ERCOT. these in a sensitivity analysis in the future.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 19 2 | Long-Term Capacity Expansion Scenarios We explore two scenarios: Current Trends TABLE 9 (CT) and Aggressive Renewables (AR). We use Key Long-Term Capacity Expansion Scenario Assumptions AURORAxmp, the Excel model, and SCM for the long-term capacity expansion analysis to obtain the Scenarios Basic Assumptions CT • Demand growth forecasts by zone: ERCOT generation portfolio in 2030. The 2030 generation (2015d) portfolio resulting from the AURORAxmp • Capital cost forecasts: ERCOT (2015e) long-term capacity expansion simulations is • PTC, ITC for new wind and solar projects used for the hourly runs (8,760 hours) using • Hardwired units under construction: 5,180 MW AURORAxmp, PLEXOS, and the Excel model. of gas including 266 MW of CHP, 4,413 MW of wind, 642 MW of solar* COMMON ASSUMPTIONS • Cost of compliance with environmental regulations (excluding CPP): ERCOT (2011) In addition to common inputs to all models • Natural gas price forecast: Hahn (2016) discussed earlier, there are some key assumptions AR • Same as CT: demand growth, capital costs, that are common across all models (Table 9). PTC/ITC, environmental compliance costs, natural gas price In the following sections, we discuss in • Additional hardwired capacity (including those greater detail key input parameters that we under development and announced): 12,106 MW of wind, 2,162 MW of solar modified or added for the long-term resource * A list of hardwired plants in two scenarios analyzed in this paper along with expansion modeling: demand forecast and their FOM, VOM and CAPEX are provided in Appendix D: ERCOT Hardwired Plant escalation; new resource candidates and their Additions for the CT Scenario and Appendix E: ERCOT Hardwired Plant Additions capital cost structures; hardwired capacities for the AR Scenario (in addition to the CT scenario). (especially for wind and solar); environmental compliance costs; and fossil fuel prices. These Demand and Energy Forecast (LTDEF): peak parameters are used across all models when load (MW) grows at an average annual rate of applicable and unless otherwise noted. 1.1% and energy load (MWh) grows at an average annual rate of 1.4% through 2025 (Figure 19). CURRENT TRENDSCurrent (CT) TrendsSCENARIO (CT) Scenario Note that both peak load and energy demand Load growth Load growth grow faster in the near future, but the growth We adopt the load growth assumptionsrate from is lower ERCOT and (2015d) flat in, knownlater years. as the Also, Long note-Term Demand and We adopt the load Energygrowth Forecast assumptions (LTDEF): from peak load (MW)that growsgrowth at rates an average have been annual much rate more of 1.1% volatile and energy load ERCOT (2015d), known(MWh) as grows the Long-Term at an average annual ratein theof 1.4% past. through Some of 2025 these (Figure variations 19). are driven Figure 19 – Year-on-Year Load Growth Rates: Historical (2007–15), Forecast (2016–25); from ERCOT (2015d) FIGURE 19 6% Year-on-Year Load Growth Rates: 5% Historical (2007–15), Forecast 4% (2016–25); from ERCOT (2015d) 3% 2% 1% 0% -1% -2% -3% -4%

Peak (MW) 2016 LTDEF Energy (GWh) 2016 LTDEF

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 20 Note that both peak load and energy demand grow faster in the near future, but the growth rate is lower and flat in later years. Also, note that growth rates have been much more volatile in the past. Some of these variations are driven by macroeconomic events such as the 2008–09 financial crisis, but weather also plays an important role. The extreme conditions in 2011 (the freeze in February, record temperatures and drought in the summer) fueled the high growth rates in 2011 and explain the contraction in 2012.

Figure 20 – ERCOT Weather Zones (ERCOT 2016d)

34

by macroeconomic events such as the 2008–09 mapped to 2003 data, which has been used as the financial crisis, but weather also plays an important representative year in recent ERCOT forecasts.15 role. The extreme conditions in 2011 (the freeze in February, record temperatures and drought However, these average growth rates vary across in the summer) fueled the high growth rates weather zones over the years. For example, in in 2011 and explain the contraction in 2012. recent years, there has been significant load growth in the Far West weather zone owing to increased FIGURE 20 drilling activity in the Permian Basin (Figure 21). ERCOT Weather Zones (ERCOT 2016d) Similarly, increased drilling activity in the Eagle Ford Shale fueled faster than historical growth in the South and partially in the South Central weather zones. Drilling activity declined significantly in both regions because of low oil prices but is expected to pick up again when oil and gas prices recover. Also, new industrial facilities such as LNG export terminals and petrochemicals facilities are under development, and their loads will have an impact primarily on the Coast and South weather zones. The LTDEF considers these factors.

We use the recent historical contribution of each region into overall energy demand (Figure 22) and peak demand (Figure 23) in ERCOT to distribute LTDEF forecasts across weather zones. We also extrapolate LTDEF forecasts through

204016 for our long-term capacity expansion The LTDEF uses actual weatherThe data LTDEF from 2002 uses –actual14 (13 years)weather to data forecast from “normal weather”runs with for AURORAxmp each of (LTDEF forecasts run the ERCOT weather zones (Figure2002–14 20) 13 (13 times. years) Each to forecast forecast “normal is ordered weather” from the highest value to the for each of the ERCOT weather zones (Figure lowest value. Then, for each ordered value, the average is calculated to obtain the “normal weather.” 20) 13 times. Each forecast is ordered from the 15 For details, please see ERCOT (2015d). These values are then mapped to 2003 data, which has been used as the representative year in recent highest value to the lowest value. Then, for each 17 16 In Figure 21 and Figure 22, we only display data through 2030 because ERCOT forecasts. ordered value, the average is calculated to obtain that is our study horizon. We run the model through 2040 to allow the the “normal weather.” These values are then model’s algorithm to calculate economics over at least 11 years of However, these average growth rates vary across weather zones over the years. For example,future cash in flows. recent years, there has been significant load growth in the Far West weather zone owing to increased drilling FIGURE 21 activity in the Permian Basin (Figure 21). 20% Historical Energy Load Growth Similarly, increased drilling activityRates byin ERCOTthe Eagle Weather Ford Zone, Shale fueled faster than historical growth in the Calculated from ERCOT (2016e) 10% South and partially in the South Central weather zones. Drilling activity declined significantly in both regions because of low oil prices but is expected to pick up again when oil and gas prices recover. Also, 0% new industrial facilities such as LNG export terminals and petrochemicals facilities are under development, and their loads will have an impact primarily on the Coast and South weather zones. The -10% LTDEF considers these factors.

We use the recent historical contribution of each region into-20% overall energy demand (Figure 22) and peak demand (Figure 23) in ERCOT to distribute LTDEF forecasts across weather zones. We also 18 extrapolate LTDEF forecasts through 2040 for our long-term-30% capacity expansion runs with AURORAxmp (LTDEF forecasts run through 2025). Since the Excel model treats ERCOT2007 as 2008a single2009 zone, it2010 only uses2011 the 2012 2013 2014 2015 aggregate demand growth figures for ERCOT. The SCM optimizes theCoast ERCOTEast systemFarWest for yearNorth 2030. North_C South South_C West

Figure 21 – Historical Energy Load Growth Rates by ERCOT Weather Zone, Calculated from ERCOT (2016e) The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 21 Figure 22 – Historical and Forecast Energy Load Growth Rates by Weather Zone 20% 17 For details, please see ERCOT (2015d). 18 In Figure 21 and Figure 22, we only display data through 2030 because10% that is our study horizon. We run the model through 2040 to allow the model’s algorithm to calculate economics over at least 11 years of future cash flows. 0%

-10% 35

-20%

-30% 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

Coast East FarWest North North_C South South_C West

The growth outlooks we use in AURORAxmp and Excel model are very close to the LTDEF projections. For ERCOT energy demand, the difference between our model inputs and ERCOT LTDEF forecasts is less than 0.9% on average. For peak load, the difference is about 2.7% for only two years and mostly less than 2% in other years, averaging 1.6%.

Figure 23 – Historical and Forecast Peak Load Growth Rates by Weather Zone

36

20%

10%

0%

-10%

-20%

-30% 2007 2008 2009 2010 2011 2012 2013 2014 2015

Coast East FarWest North North_C South South_C West

Figure 22 – Historical and Forecast Energy Load Growth Rates by Weather Zone FIGURE 22 20% Historical and Forecast Energy Load Growth Rates 10% by Weather Zone

0%

-10%

-20%

-30% 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

Coast East FarWest North North_C South South_C West

The growth outlooks20% we use in AURORAxmp and Excel model are very close to the LTDEF projections. FIGURE 23 20% For ERCOT energy demand, the difference between our model inputs and ERCOT LTDEF forecasts is less Historical and Forecast Peak Load 10% than 0.9% on average.10% For peak load, the difference is about 2.7% for only two years and mostly less Growth Rates by Weather Zone than 2% in other0% years, averaging 1.6%. 0% -10% Figure 23 – Historical and Forecast Peak Load Growth Rates by Weather Zone -10% -20% -20% -30% -30% -40% -40% 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 Coast East FarWest North 36 Coast East FarWest North North_C South South_C West North_C South South_C West

Figure 24 – Annual Energy Demand Growth Rates in Four Main AURORAxmp Zones for ERCOT Figure 24 – Annual Energy Demand Growth Rates in Four Main AURORAxmp Zones for ERCOT FIGURE 24 7% 7% Annual Energy Demand Growth Rates in Four Main 6% 6% AURORAxmp Zones for ERCOT 5% 5% 4% 4% 3% 3% 2% 2% 1% 1% 0% 0% 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 ERCOT-Houston ERCOT-North ERCOT-South ERCOT-West ERCOT-Houston ERCOT-North ERCOT-South ERCOT-West

Regional forecasts used by AURORAxmp are also consistent with the LTDEF forecasts, and, importantly, captureRegional the forecasts expectation used byof AURORAxmpfaster growth arein the also Far consistent West (or with West the in AURORAxmpLTDEF forecasts, terminology) and, importantly, relative The Full Cost of Electricitytocapture other (FCe-) thezones expectation (Figure 24 of). However,faster growth the eightinModel the load DocumentationFar Westzones (or inand AURORAxmp WestResults forin ERCOTAURORAxmp Scenarios,do not Julyperfectly terminology) 2017 | 22 correspond relative to ERCOTto other weather zones ( Figurezones. 1924 We). However, matched the the eight grow loadth rates zones in thein AURORAxmp model as closely do not as perfectlypossible tocorrespond the LTDEF to 19 projectionsERCOT weather. This zones. regional We treatment matched of the load grow growthth rates in AURORAxmp in the model, comparedas closely as to possiblethe single to- zonethe LTDEF treatmentprojections in. This the regionalExcel model treatment and SCM of ,load is one growth of the in factors AURORAxmp that could, compared lead to differences to the single in- zoneresults. treatment in the Excel model and SCM, is one of the factors that could lead to differences in results.

19 The four main zones in AURORAxmp are aggregations of the eight zones used by ERCOT; AURORAxmp also has 19 four The relatively four main small zones zones in AURORAxmp for Austin Energy, are aggregations CPS, LCRA and of the Rayburn eight zonesterritories. used by ERCOT; AURORAxmp also has four relatively small zones for Austin Energy, CPS, LCRA and Rayburn territories. 37 37

TABLE 10 Overnight Capital Expenditures by Generation Technology [$2015/kW], adapted from ERCOT (2015e)

Year CC CT Coal Nuclear IGCC Wind Solar PV Biomass Geo Battery CAES 2015 1,073 791 3,202 6,395 4,307 1,740 1,781 3,903 5,025 718 1,051 2016 1,073 791 3,202 6,395 4,307 1,682 1,541 3,903 5,025 677 1,044 2017 1,073 791 3,202 6,395 4,307 1,627 1,352 3,903 5,025 633 1,039 2018 1,073 791 3,202 6,395 4,307 1,573 1,241 3,903 5,025 575 1,033 2019 1,073 791 3,202 6,395 4,307 1,521 1,191 3,903 5,025 543 1,028 2020 1,073 791 3,202 6,395 4,307 1,477 1,149 3,903 5,025 512 1,022 2021 1,073 791 3,202 6,395 4,307 1,436 1,110 3,903 5,025 496 1,016 2022 1,073 791 3,202 6,395 4,307 1,395 1,074 3,903 5,025 480 1,011 2023 1,073 791 3,202 6,395 4,307 1,355 1,045 3,903 5,025 465 1,005 2024 1,073 791 3,202 6,395 4,307 1,317 1,024 3,903 5,025 463 999 2025 1,073 791 3,202 6,395 4,307 1,280 1,005 3,903 5,025 461 995 2026 1,073 791 3,202 6,395 4,307 1,253 986 3,903 5,025 458 989 2027 1,073 791 3,202 6,395 4,307 1,226 968 3,903 5,025 456 983 2028 1,073 791 3,202 6,395 4,307 1,201 950 3,903 5,025 453 978 2029 1,073 791 3,202 6,395 4,307 1,176 932 3,903 5,025 451 973 2030 1,073 791 3,202 6,395 4,307 1,151 915 3,903 5,025 448 967 2031 1,073 791 3,202 6,395 4,307 1,127 898 3,903 5,025 446 962 through 2025). Since the Excel model treats model as closely as possible to the LTDEF ERCOT as a single zone, it only uses the aggregate projections. This regional treatment of load growth demand growth figures for ERCOT. The SCM in AURORAxmp, compared to the single-zone optimizes the ERCOT system for year 2030. treatment in the Excel model and SCM, is one of the factors that could lead to differences in results. The growth outlooks we use in AURORAxmp and Excel model are very close to the LTDEF Capital costs projections. For ERCOT energy demand, the difference between our model inputs and For the long-term capacity expansion analysis, ERCOT LTDEF forecasts is less than 0.9% an important input is the overnight capital on average. For peak load, the difference is expenditures for building a new plant. There are about 2.7% for only two years and mostly less various sources for such data such as Black & than 2% in other years, averaging 1.6%. Veatch (2012), EIA (2013), and Lazard (2014). In addition, there are forecasts from industry Regional forecasts used by AURORAxmp are associations and reports from the Lawrence also consistent with the LTDEF forecasts, and, Berkeley National Laboratory on past cost importantly, capture the expectation of faster performance. There is no consensus on future growth in the Far West (or West in AURORAxmp forecasts. Even the past data indicate significant terminology) relative to other zones (Figure 24). regional variability. Updates are provided regularly However, the eight load zones in AURORAxmp because of technological improvements and the do not perfectly correspond to ERCOT weather economies of scale resulting from the expansion 17 zones. We matched the growth rates in the of installed capacity of newer technologies.

Although the cost estimates for conventional 17 The four main zones in AURORAxmp are aggregations of the eight zones used by ERCOT; AURORAxmp also has four relatively small zones for technologies such as combined cycle and Austin Energy, CPS, LCRA and Rayburn territories. combustion turbines are fairly well established,

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 23 Figure 25 – Annual Decline Rates in Overnight Capital Expenditures of Wind, Solar, Battery Storage, CAES

FIGURE 25 18% Annual Decline Rates in Overnight 16% Capital Expenditures of Wind, 14% Solar, Battery Storage, CAES* 12% 10% 8% 6% 4% 2% 0%

Wind Solar Battery CAES * Currently, AURORAxmp long-term resource expansion algorithm does not include either battery or CAES as new resource candidates. Therefore the algorithm will not decide to build these energy storage resources although we can hardwire storage.

there are regional variations even for those, and and solar costs in real terms decline roughly technological improvement cannot be ruled out 2.7% and 4.9% per year but faster in the near (e.g., current combined cycle plants are more future (Figure 25). In contrast, the EIA assumes efficient and have better ramping capabilities). about $2,600/kW for solar PV in 2015 as More uncertain are the expectations on capital compared to about $1,800/kW by ERCOT.19 expenditures of wind and solar as well as other technologies with currently limited market share For the long-term capacity expansion simulation (e.g., storage). On the other hand, in times of with AURORAxmp, we convert the capital cost high economic growth, costs of capital-intensive ($/kW) shown in Table 10 into Base Capital projects such as power plants are likely to increase Carrying Cost in $/MW-week, an annuitized driven by increased competition for engineering, fixed payment after accounting for assumed procurement and construction services as well as tax rates, depreciation schedule, book life for steel, specialized equipment, and other supplies. various generation types, capital structure, and costs of debt and equity (Table 11). Given these complexities, it is possible for different analysts to develop divergent forecasts for capital For example, advanced CC gas units have costs of different technologies. In fact, although overnight capital cost of $1,073/kW (Table 10). keeping overnight capital expenditures for Utilizing assumptions in Table 11, we convert conventional technologies the same over the years $1,073/kW overnight cost into an annuitized in real terms18 seems to be commonly (though fixed Base Capital Carrying Cost of $1,693 / not universally) accepted, there are a variety of MW-week. If the model decides to build a new assumptions on emerging technologies. Since we Advanced CC unit in year 2016, then this unit wanted to maintain our results’ comparability will incur fixed Base Capital Carrying Cost for to scenarios developed by ERCOT, we used the every year from 2016–35 (20-year book life), in ERCOT LTSA 2016 assumptions (Table 10). addition to any fixed and variable O&M costs.

These estimates imply the same cost every year in The most significant differences between the real terms for conventional thermal technologies parameters assigned to generation technologies as well as IGCC and biomass. In contrast, wind

19 AEO 2016 cost assumptions for other technologies, which are based on 18 Real costs here mean adjusted for inflation. an engineering analyses, are also somewhat different: see EIA (2016e).

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 24 TABLE 11 Financial Parameters for Calculating Base Capital Carrying Cost for Potential New Resources

Adv. Adv. CC Pulv. Solar Solar Adv. CT CT Biomass Geothermal IGCC Nuclear Wind CC 1x1 Coal Thermal PV

Debt Return 7.5% 7.5% 7.5% 7.5% 7.5% 7.5% 4.8% 4.8% 4.8% 12.0% 12.0% 8.0% Equity Return 9.0% 9.0% 9.0% 9.0% 9.0% 9.0% 8.0% 8.0% 8.0% 7.0% 7.0% 9.6% Debt % 65% 65% 40% 40% 40% 40% 40% 40% 40% 60% 60% 55% Equity % 35% 35% 60% 60% 60% 60% 60% 60% 60% 40% 40% 45% Composite 8.0% 8.0% 8.4% 8.4% 8.4% 8.4% 6.7% 6.7% 6.7% 10.0% 10.0% 8.7% Cost Interest 1.7% 1.7% 1.1% 1.1% 1.1% 1.1% 0.7% 0.7% 0.7% 2.5% 2.5% 1.5% Deduction After Tax Cost 6.32% 6.32% 7.35% 7.35% 7.35% 7.35% 6.04% 6.04% 6.04% 7.48% 7.48% 7.18% of Capital Real After Tax 3.73% 3.73% 4.73% 4.73% 4.73% 4.73% 3.45% 3.45% 3.45% 4.86% 4.86% 4.57% Cost of Capital Book Life Years 20 20 20 20 20 20 30 30 30 20 20 20 Tax Recovery 20 20 15 15 5 5 20 15 20 5 5 5 Period Notes: (1) We assume 2.5% for general inflation and 35% Federal Business Income Tax; in Texas, there is no state income tax. (2) The Tax Recovery Period is based on IRS Modified Accelerated Cost Recovery System (MACRS). (3) For debt/equity returns and ratios, we refer to (i) Washington State Department of Revenue (2016); (ii) NREL (2013); and (iii) NREL (2014).

are the capital structure and costs of debt and law, for wind facilities that have begun commercial equity. Although we tried to reflect these financing operation after 2006, we added a negative adder terms as realistically as possible, these terms are ($/MWh) to its variable cost in a nested structure. often project-specific and data are not always Prior to 2018, each wind facility had a production publicly available. As such, it is possible for these tax credit of $23/MWh for 10 years, which will be assumptions to contribute to differences between reduced to $18.40, $13.80, and $9.20 per year for our results and those of ERCOT, among other 10 years in 2018, 2019 and 2020 respectively. For analyses. Still, their impact on capital carrying costs solar facilities, we incorporated ITC similarly as is less significant than the impact of the federal a negative cost adder, but this time as capital cost subsidies (PTC and ITC). The Excel model uses the converted to $/MW-week basis and included in same underlying discount rates as other approaches. the fixed-cost input of solar resources (Table 12).

Tax credits TABLE 12

The federal production and investment tax credits Levelized Investment Tax Credit for Solar PV Facilities [$/MW-Week) (PTC and ITC) are applied to wind and solar as outlined in the Consolidated Appropriations Act, Year 2015 2016 2017 2018 2019 2020 2021 2022 201620 in both AURORAxmp and Excel models. Levelized -596 -516 -453 -416 -399 -333 -272 -120 The Act extended the expiration date for PTC and ITC ITC, with a phase-down schedule.21 To reflect the Note: The decline trend of Levelized Investment Tax Credit reflects the combination of overnight cost improvement and phase-down of ITC.

20 See United States Congress (2015).

21 See Renewable Electricity Production Tax Credit (PTC) at DOE (2016); Federal Investment Tax Credit (ITC) at DSIRE (2016).

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 25 Hardwired new builds and retirements We also hardwire the retirement of roughly 840 MW of coal plants at the announced date There are several power plants in ERCOT that of mothballing. Otherwise, the model uses its are already under construction or in advanced economic logic to decide whether to retire a unit. stages of development. The model is not likely Wind turbines are said to have an economic life to capture these capacities as its logic requires of 25 years. However, industry news suggests a multi-year forward look. We also do not want that most wind turbines will be retrofitted the model to build extra capacity because price before reaching that age and will continue to signals are misled by the absence of these units operate. Ideally, we would want to introduce a in our resources database. This is particularly cost-adder for this retrofit similar to cost adders an issue for renewables. AURORAxmp typically to comply with environmental regulations does not build as much renewables capacity as discussed in the next section. However, we do is under construction; when the model builds not have data on the cost of wind turbine retrofits any wind or solar capacity, it occurs in late at this time. We decided to extend the life of 2020s when the overnight CAPEX of these wind capacity beyond 25 years to keep them technologies are expected to be lower than that operational throughout our study horizon. of gas units. Similarly, the Excel model adds 1.5 GW of Coastal Wind in 2029, 2 GW of Inland Environmental regulations compliance Wind in 2030, and 1.5 GW of PV in 2029 in the absence of assumed hardwired constructions. There are a series of environmental regulations that threaten the retirement of some thermal units. We have seen this result repeatedly over the last Outside ERCOT, the threat of these environmental five years of running the AURORAxmp model. regulations has already caused many retirements The main reason appears to be that we cannot (primarily coal but also some older gas units). This consistently capture long-term power purchase threat was made stronger by shrinking revenue agreements that are granted to wind and solar margins driven by low natural gas prices and projects by utilities and cooperatives, local tax and increasing share of renewables dispatched at low other benefits offered by municipalities, or revenues or sometimes negative prices owing to their near- from the sale of Renewable Energy Certificates zero marginal costs and PTC revenues. Shrinking (the latter has not been significant in Texas for a revenues seem to be the main reason for some long time but it matters in other jurisdictions). early nuclear retirements as well. We do not consider the impacts of the Clean Power Plan. As such, we hardwire those projects that are already under construction in the CT scenario, TABLE 13 including roughly 4,900 MW of natural gas Retrofit Costs for Coal Power Plants ($/MW-Week). units, 640 MW of utility scale solar, and 4,400 Retrofit Cost Fixed Cost Adder ($/MW-Week) 22 MW of wind resources (Table 9). For thermal ($/kW) from 2015 to 2024 resources, having a signed interconnection 50 115 agreement with ERCOT is sufficient as this 200 460 process requires developers to meet numerous 300 690 technical criteria and acquire all environmental 400 920 permits in addition to demonstration of access 450 1,035 23 to water (e.g., acquisition of water rights). 650 1,495 700 1,610 22 The following sources were used for projects under construction and future projects: ERCOT (2015c), ERCOT (2016b), EIA (2016b), EIA ERCOT (2011) provides cost estimates for coal and (2016c), PUCT (2015), SNL (2016), SEIA (2016), TCEQ (2015), TCEQ gas units in ERCOT to comply with Clean Water (2016), and TCPA (2016). Operational details of these units are provided in Appendix E. (storage), CCGT: 16,589 MW, CCGT+CCS: 500 MW, Hydro: 10 MW, IC: 23 We excluded some projects because they did not meet any of our 94 MW (NG, biomass), IGCC+CCS: 240 MW (coal), OCGT: 9,666 MW, ST: selection criteria or were insignificant capacity: CAES: 911 MW 5,010 MW (nuclear).

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 26 TABLE 14 Act Section 316(b), coal combustion residuals disposition (coal ash), Clean Air Act – Hazardous Retrofit Costs Assignment based on ERCOT (2011) Air Pollutants (HAP), and Clean Air Transport Rule Name Retrofit Cost Assignment [$/kW] (CATR). Although not identical laws, we treat HAP Big Brown #1 650 compliance as a substitute for Mercury and Air Big Brown #2 650 Toxics Standards (MATS) and CATR as a substitute Coleto Creek #1 50 for Cross-State Air Pollution Rule (CSAPR). Fayette Power Prj #1 400 Also, based on ERCOT (2014), we assume that Fayette Power Prj #2 400 scrubber costs provided in ERCOT (2011) also Fayette Power Prj #3 450 provide compliance with the Regional Haze rule. Gibbons Creek #1 300 Accordingly, we first estimate the potential retrofit JK Spruce #1 200 cost ($/kW) for coal units to comply with these JT Deely #1 50 environmental regulations. We then convert JT Deely #2 50 these costs into a fixed-cost adder in $/MW-week Limestone #1 200 basis between 2015–24 (Table 13). The estimated Limestone #2 200 retrofit costs of individual coal units in ERCOT are Martin Lake #2 400 reported in Table 14. Since we impose these retrofit Martin Lake #3 400 costs, we run the model without emissions costs for Monticello #1 700 Monticello #2 700 SO2 and NOX to avoid penalizing these plants twice. Monticello #3 400 Candidates for new thermal builds are assumed Oklaunion #1 200 to be compliant with these new regulations. These San Miguel #1 300 costs are considered only in the AURORAxmp Sandow #4 200 runs. However, units with the highest retrofit costs WA Parish #5 50 are not the units that retire in the AR scenario WA Parish #6 50 contrary to what one might expect. Still, most of the WA Parish #7 50 units that retire have retrofit costs of $200-300/kW. Martin Lake #1 400 Fuel Price Forecasts higher than any of the other more recent forecasts The natural gas price is the most important fuel and lacks cyclicality (Figure 26). In our analysis, we price in ERCOT given that more than half of use the mean Henry Hub natural gas price forecast generation has been from gas-fired power plants produced via statistical modeling by Hahn (2016) in recent history. More importantly, natural gas rather than any of these alternative forecasts. fired generation is often the marginal supplier of electricity. Coal prices (both subbituminous In AURORAxmp, we impose monthly natural gas and lignite) have been much more stable than shapes from the IHS forecast on annual natural gas natural gas prices. Given the volatile history price forecast generated by Hahn (2016), and then of natural gas prices, the current turmoil in extrapolate monthly prices through 2040 to input the oil and gas industry, and the uncertainty into the model, which also has basis differentials associated with natural gas demand from various across different ERCOT zones. The Excel model sectors, there are many forecasts available. uses the annual averages. The large differences between the forecast used by the ERCOT LTSA ERCOT uses an average forecast for Henry Hub and forecast by Hahn (2016) are likely to have a natural gas price from EIA scenarios (Reference and significant impact on capacity expansion runs. High Oil & Gas Resource) employed in the Annual Given that basis differentials across ERCOT are Energy Outlook 2015.24 This forecast is significantly

24 The average of natural gas price forecasts from AEO 2016 Reference the AEO 2015 estimates, with larger differences in the early years of the and High Oil & Gas Resource scenarios is on average $0.38 lower than forecast.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 27 mean Henry Hub natural gas price forecast produced via statistical modeling by Hahn (2016) rather than any of these alternative forecasts.

In AURORAxmp, we impose monthly natural gas shapes from the IHS forecast on annual natural gas price forecast generated by Hahn (2016), and then extrapolate monthly prices through 2040 to input into the model, which also has basis differentials across different ERCOT zones. The Excel model uses the annual averages. The large differences between the forecast used by the ERCOT LTSA and forecast by Hahn (2016) are likely to have a significant impact on capacity expansion runs. Given that basis differentials across ERCOT are small, the way natural gas price forecasts are captured in AURORAxmp and the Excel model are not likely to cause significant discrepancy in results.

Figure 26 – Henry Hub Natural Gas Price Forecasts [$2015/MMBtu] FIGURE 26 $5.0 Henry Hub Natural Gas Price $4.5 Forecasts [$2015/MMBtu] $4.0 $3.5 $3.0 $2.5 $2.0 $/MMBtu $1.5 $1.0 $0.5 $0.0

ERCOT 2016 LTSA Hahn (2016) IHS SNL WM Base WM $65/bbl WM $50/bbl

For coal prices, we first retrieved historical prices (2011–15) for lignite and subbituminous coal from EIA Form 923 Schedule 2: Fuel Receipts and Costs. We then took a five-year rolling average and extrapolated out to 2040 (Figure 27). Accordingly, we also assume coal units in Texas use either lignite or FIGURE 27 subbituminous coal$3.0 based on EIA Form 923 Schedule 3 generator data (Appendix A: ERCOT Plant Coal Price Forecasts [$2015/MMBtu]Database for End-of-Year 2015). $2.5 Figure 27 – Coal Price Forecasts [$2015/MMBtu] $2.0

$1.5 $/MMBtu $1.0

$0.5

$0.0 2015 2020 2025 2030

US Coal Lig US Coal Sub

44 Aggressive Renewables (AR) Scenario small, the way naturalAll gasof the price assumptions forecasts are and inputs describedAGGRESSIVE above for RENEWABLES the CT scenario also (AR) apply SCENARIO in the AR scenario. The captured in AURORAxmponly difference and the isExcel the additional model are hardwired renewables capacity: more than 12 GW of wind and a little All of the assumptions and inputs described above not likely to cause significantover 2 GW ofdiscrepancy solar projects. in results. These projects are either recently announced, or in various stages of for the CT scenario also apply in the AR scenario. development (Appendix E: ERCOT Hard-Wired Plant Additions for the AR Scenario (in addition to the CT For coal prices, we first retrieved historical prices The only difference is the additional hardwired scenario)). (2011–15) for lignite and subbituminous coal renewables capacity: more than 12 GW of wind from EIA Form 923 Schedule 2: Fuel Receipts and and a little over 2 GW of solar projects. These Costs. We then tookLong a five-year-Term rolling Capacity average andExpansion projects Resultsare either recently announced, or in extrapolated out to 2040Although (Figure we use27). the Accordingly, same set of inputvarious data andstages assumptions of development as much (Appendix as possible, E). the three models we also assume coalused units for in longTexas-term use capacityeither expansion simulations are structurally different as discussed earlier in this lignite or subbituminousreport. coal They based also onyield EIA different Form outputs in addition to common outputs. As such, we discuss the results 923 Schedule 3 generatorby model data first (Appendix before providing A). a comparison. AURORAxmp New Builds and Retirements The Full Cost of ElectricityUnder (FCe-) the AR scenario, a significant portionModel of Documentation coal capacity and Results is retired for ERCOT between Scenarios, July2015 2017–30 | 28(Table 15). Given that we roughly quadrupled each of wind and solar hard-wired capacities in this scenario, this result is not surprising. However, in the CT scenario, only 840 MW of coal capacity is retired, despite the environmental retrofit costs discussed earlier. The model does not retire any natural gas units in either scenario, potentially owing to the low natural gas price forecast. The model builds no new wind and only 450 MW of solar in the AR scenario beyond the hard-wired wind and solar capacities.

Table 15 – Summary of Resources under the Two Scenarios (MW) with AURORAxmp Coal NG NG Wind Solar NG Wind Solar Net Total Retire Retire Hardwire Hardwire Hardwire New New New Additions Installed in Build Build Build 2015-30 2030 CT 840 0 5,180 4,413 642 4,690 -- -- 14,085 106,794 AR 6,433 0 5,180 16,519 2,800 6,360 -- 450 24,876 117,585

45

3 | Long-Term Capacity Expansion Results

Although we use the same set of input data and Annual Capacity Changes, Average assumptions as much as possible, the three models Wholesale Prices and Reserve Margins used for long-term capacity expansion simulations are structurally different as discussed earlier. All of the coal retirements in both scenarios They also yield different outputs in addition to happen by 2019 with the exception of 600 MW common outputs. As such, we discuss the results retired in 2023 in the AR scenario (Figure 28), by model first before providing a comparison. probably in response to hardwired gas, wind and solar units leading to relatively low prices AURORAxmp and relatively high reserve margins in the early years of the study horizon (Figure 29). New Builds and Retirements Natural gas price remains below $3 in real terms Under the AR scenario, a significant portion of throughout the study period (2030) (Figure 26), coal capacity is retired between 2015–30 (Table 15). helping to suppress energy market prices along Given that we roughly quadrupled each of wind with near-zero marginal cost wind and solar and solar hardwired capacities in this scenario, capacity. There are negative price periods as well this result is not surprising. However, in the CT owing to PTC credits received by wind farms, scenario, only 840 MW of coal capacity is retired, especially in the AR scenario. For example, prices despite the environmental retrofit costs discussed were negative in 7% of the hours in 2016 in the earlier. The model does not retire any natural gas West zone, and 5-6% of the hours in other zones units in either scenario, potentially owing to the low except for the Houston zone, where prices were natural gas price forecast. The model builds no new almost always positive. In 2017, there were more wind and only 450 MW of solar in the AR scenario hours with negative prices: 9% in the West zone, beyond the hardwired wind and solar capacities. and 6-7% in other zones (except the Houston zone). Even in 2018 and 2019, 3-4% of prices Because of low to moderate coal capacity were negative in all zones except Houston. The retirements and hardwired renewable additional cost of emission control equipment capacities, new gas-fired capacity builds (Table 13) probably undermined the economics remain relatively low given low gas prices: of these plants in this low-price environment. about 4,700 MW in the CT scenario and about 6,400 MW in the AR scenario, all of which are These conditions encourage new advanced CCs in advanced CC units in the Houston area. several-year cycles, especially in the CT scenario.

TABLE 15 Summary of Resources under the Two Scenarios (MW) with AURORAxmp

NG Wind Solar Net Total Coal NG NG Wind Solar New New New Additions Installed Retire Retire Hardwire Hardwire Hardwire Build Build Build 2015-30 in 2030

CT 840 0 5,180 4,413 642 4,690 -- -- 14,085 106,794

AR 6,433 0 5,180 16,519 2,800 6,360 -- 450 24,876 117,585

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 29 There are fewer negative prices under the CT million MWh (about 4%). There is more solar scenario (2-3% in 2016 and about 1% in 2017); coal generation (about 48 million MWh) in the AR retirements are limited in this scenario. scenario. However, solar generation is relatively small: 73 million MWh in the AR scenario. The average prices are roughly the same across the two scenarios until the mid-2020s after which System Costs they are higher in the CT scenario (Figure 29). This can be expected given the higher share of The comparison of total system costs from 2015 to low-cost renewables in the AR scenario. Also, 2030 across the two scenarios is informative. Total reserve margins are higher in the AR scenario system costs include fixed and variable O&M costs, in many years. Given that peak wind generation base capital carrying costs of new plants built by in ERCOT does not typically overlap with peak the model (i.e., overnight capital cost distributed load hours, this result might appear surprising at over the life of the plant), and fuel costs of all first. However, hardwired wind capacity is large resources. Total system costs are slightly larger and some of it is in coastal areas where wind (about $10 billion) in the AR scenario (Figure 31). generation is more coincident with peak load. Although the capital costs for wind and solar Total Generation decline over time (Table 10), they remain more The main difference between the two scenarios expensive than gas units through the early 2020s, in terms of generation concerns wind and coal. when most hardwired capacity were added to the There is more wind generation to replace primarily resources database. As a result, capital costs in coal and some natural gas generation in the the AR scenario add up to almost $40 billion as AR scenario as compared to the CT scenario compared to about $19 billion in the CT scenario. (Figure 30). Coal generation is reduced by 495 The associated fixed costs are also higher by about million MWh (about 31%). Wind generation $3 billion. However, there are significant fuel cost increases by about 581 million MWh (about savings in the AR scenario (roughly $12 billion) 49%) and natural gas generation declines by 114 and operating cost savings (about $3.5 billion).

FIGURE 28 Annual Capacity Additions and Figure 28 – RetirementsAnnual Capacity with Additions AURORAxmp and Retirements with AURORAxmp

14,000 14,000 CT AR 12,000 12,000

10,000 10,000

8,000 8,000

6,000 6,000 MW MW 4,000 4,000

2,000 2,000

0 0

‐2,000 ‐2,000

Coal Retire NG Retire NG Hardwire Wind Hardwire Coal Retire NG Retire NG Hardwire Wind Hardwire Sun Hardwire NG NewBuild Wind New Build Sun New Build Sun Hardwire NG NewBuild Wind New Build Sun New Build

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 30 FIGURE 29 FigureFigure 29 29 – – Annual Annual Average Average Prices Prices and and Reserve Reserve Margins Margins with with AURORAxmp AURORAxmp Annual Average Prices and Reserve Margins with AURORAxmp

$60$60 CTCT 25%25% $60$60 ARAR 25%25%

$50$50 $50$50 20%20% 20%20%

$40$40 $40$40 15%15% 15%15% Margin Margin Margin Margin $30$30 $30$30 ($2015/MWh) ($2015/MWh) ($2015/MWh) ($2015/MWh)

10%10% 10%10%

$20$20 Total Generation Reserve Reserve $20$20 Reserve Reserve Price Price Price Price The main difference between5%5% the two scenarios in terms of generation concerns wind and5%5% coal. There is $10$10 $10$10 more wind generation to replace primarily coal and some natural gas generation in the AR scenario as $0$0 compared to the CT scenario0%0% (Figure$0$0 30). Coal generation is reduced by 495 million MWh0%0% (about 31%). Wind generation increases by about 581 million MWh (about 49%) and natural gas generation declines by 114 million MWh (about 4%). There is more solar generation (about 48 million MWh) in the AR RMRM PricePrice RMRM PricePrice scenario. However, solar generation is relatively small: 73 million MWh in the AR scenario.

Figure 30 – Total Generation Output by Fuel Type from 2015 to 2030 with AURORAxmp FIGURE 30 7,000 Total Generation Output by Fuel Type from 2015 to 2030 with AURORAxmp 6,000 Wind 1,184 Wind 5,000 1,765

4,000 Natural Gas 2,741 Natural Gas 3,000 2,627 Million MWh Million 2,000 Coal Coal 1,620 1,000 1,125 Nuclear, 614 Nuclear, 603 0 Current Trends Aggressive Renewables

Other Hydro Nuclear Coal Natural Gas Wind Solar

System Costs FIGURE 31 $200 The comparison of total system costs from 2015 to 2030 across the two scenarios is informative. Total Total System Costs from 2015 $180 to 2030 with AURORAxmp system costs include fixed and variable O&M costs, base capital carrying costs of new plants built by the $160 model (i.e., overnight capital cost distributed over the life of the plant), and80.6 fuel costs of all resources. Total system costs $140are slightly larger (about92.6 $10 billion) in the AR scenario (Figure 31). $120 Although the capital costs for wind and solar decline over time (Table 10), they remain more expensive $100 18.6 than gas units through the early 2020s, when most hard-wired capacity were added to the resources $80 22.0 database. As a res $2015Billion ult, capital costs in the AR scenario add up to almost $4054.4 billion as compared to about $19 billion in the CT $60scenario. The associated fixed costs are also higher by about $3 billion. However, 51.4 there are significant$40 fuel cost savings in the AR scenario (roughly $12 billion) and operating cost savings (about $3.5 billion). $20 40.3 18.7 $0 Figure 31 – Total System Costs from 2015 to 2030 with AURORAxmp Current Trends Aggressive Renewables

Base Capital Carrying Fixed O&M Variable O&M Fuel

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 31 Excel Model New Builds and Retirements Under the CT scenario, the Excel model makes minor adjustments beyond the assumed hardwired plants: only 500 MW of Coastal Wind are added (Table 16). Under the AR scenario, the Excel model does 48 not make any additions or retirements beyond hardwires. By comparison, the primary difference is that

AURORAxmp expanded gas capacity by 4.7 GW and 6.4 GW under the CT and AR scenarios, respectively. This gas build-out contributes to the higher installed 2030 capacities in AURORAxmp compared to Excel, especially in the CT scenario. The difference is negligible in the AR scenario because gas additions in AURORAxmp mostly compensate for coal retirements, which do not occur with the Excel model. As a reminder, the Excel model did not include coal retrofit costs which may contribute to less coal retirement and natural gas expansion.

Table 16 – Summary of Resources under the Two Scenarios (MW) with the Excel model Coal NG NG Wind Solar NG Wind Solar Net Total Retire Retire Hardwire Hardwire Hardwire New New New Additions Installed in hardwire Build Build Build 2015-30 2030 CT 840 -- 5,180 4,413 642 -- 500 9,895 103,332 AR 840 -- 5,180 16,519 2,800 ------23,660 117,097

Annual Capacity Changes, Average Wholesale Prices and Reserve Margins Unlike AURORAxmp, the Excel model does not retire any more coal units than the hardwired 840 MWs in the AR scenario. Partially because of the lack of coal retirements, there are no new gas builds. However, the Excel model does not build any new gas capacity even in the CT scenario although both AURORAxmp and Excel only have the 840-MW hardwired coal retirement in this scenario. The only other difference is the 500 MW of new wind capacity built in the CT scenario with the Excel model.

Annual average reserve margins with the Excel model are highly correlated with those from AURORAxmp (comparing Figure 33 to Figure 29): 0.87 in the CT scenario and 0.8 in the AR scenario. The

49

Excel Model AURORAxmp (comparing Figure 33 to Figure 29): 0.87 in the CT scenario and 0.8 in the AR New Builds and Retirements scenario. The average over the study horizon is the same in the CT scenario across the two models; Under the CT scenario, the Excel model makes but it is almost 3% higher with the Excel model minor adjustments beyond the assumed hardwired in the AR scenario. A possible explanation is that plants: only 500 MW of Coastal Wind are added AURORAxmp calculates the average ERCOT (Table 16). Under the AR scenario, the Excel reserve margin internally, taking into account model does not make any additions or retirements differences (e.g., wind and solar shapes) across eight beyond hardwires. By comparison, the primary zones and plant availabilities. In the Excel model difference is that AURORAxmp expanded gas an ex-post calculation is conducted for the whole capacity by 4.7 GW and 6.4 GW under the CT of ERCOT with nameplate capacities for thermal and AR scenarios, respectively. This gas build- out contributes to the higher installed 2030 units, a single wind shape, and a single solar shape. capacities in AURORAxmp compared to Excel, Average prices smoothly rise in both scenarios especially in the CT scenario. The difference is with the Excel model. Although the overall upward negligible in the AR scenario because gas additions trend after 2017 is consistent with the average in AURORAxmp mostly compensate for coal prices obtained from AURORAxmp runs, the retirements, which do not occur with the Excel correlation between price series from two model model. As a reminder, the Excel model did not runs is only moderately correlated in either include coal retrofit costs which may contribute to scenario (0.55). The prices follow a more cyclical less coal retirement and natural gas expansion. pattern in the AURORAxmp runs. Again, ERCOT Annual Capacity Changes, Average prices reported in Figure 29 are averages of hourly Wholesale Prices and Reserve Margins zonal prices internally calculated by AURORAxmp whereas an ex-post calculation is conducted for Unlike AURORAxmp, the Excel model does not all of ERCOT in the Excel model. Given that gas retire any more coal units than the hardwired is almost always on the margin, especially in the 840 MWs in the AR scenario. Partially because CT scenario, average ERCOT prices increase at of the lack of coal retirements, there are no the same pace as the price of natural gas (Figure new gas builds. However, the Excel model 26). The correlation between average wholesale does not build any new gas capacity even in electricity prices and natural gas prices is 0.95 in the CT scenario although both AURORAxmp the CT scenario and 0.89 in the AR scenario. and Excel only have the 840-MW hardwired coal retirement in this scenario. The only other Total Generation difference is the 500 MW of new wind capacity built in the CT scenario with the Excel model. The Excel model yields very similar results to those from AURORAxmp, especially in the AR scenario. Annual average reserve margins with the Excel In CT scenario, the Excel model substitutes model are highly correlated with those from about 200 million MWh of coal with gas.

TABLE 16 Summary of Resources under the Two Scenarios (MW) with the Excel model

Coal Wind Solar Net Total NG NG Wind Solar NG Retire New New Additions Installed Retire Hardwire Hardwire Hardwire New Build Hardwire Build Build 2015-30 in 2030

CT 840 -- 5,180 4,413 642 -- 500 9,895 103,332

AR 840 -- 5,180 16,519 2,800 ------23,660 117,097

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 32 FIGURE 32 Annual Capacity Additions and Retirements with the Excel model FigureFigure 32 32 – –Annual Annual Capacity Capacity Additions Additions and and Retirements Retirements with with the the Excel Excel model model

14,00014,000 14,00014,000 CTCT ARAR 12,00012,000 12,00012,000

10,00010,000 10,00010,000

8,0008,000 8,0008,000

6,0006,000 6,0006,000 MW MW MW MW 4,0004,000 4,0004,000

2,0002,000 2,0002,000

00 00

‐2,000‐2,000 ‐2,000‐2,000

CoalCoal Retire Retire NGNG Retire Retire NGNG Hardwire Hardwire WindWind Hardwire Hardwire CoalCoal Retire Retire NGNG Retire Retire NGNG Hardwire Hardwire WindWind Hardwire Hardwire SunSun Hardwire Hardwire NGNG NewBuild NewBuild WindWind New New Build Build SunSun New New Build Build SunSun Hardwire Hardwire NGNG NewBuild NewBuild WindWind New New Build Build SunSun New New Build Build

System Costs model, whereas the increase is only 5% with AURORAxmp. Cost estimates for each category System costs with the Excel model are consistent in the AR scenario are closer to each other (3–9%) with those from AURORAxmp (Figure 35). Total across the two models but there is a significant costs are almost the same with the CT scenario and difference in base capital carrying cost in the CT within 3% with the AR scenario. With both models, scenario: $14.6 billion in the Excel model versus the shares of different cost categories are roughly $18.7 billion in AURORAxmp (28% difference). the same in each scenario. Fuel costs account for This difference can be explained by the cost of the about half of the total costs in the CT scenario but 4,700-MW new gas builds with AURORAxmp; only 42% in the AR scenario. Despite the fuel cost the Excel model does not build any new units savings, capital costs more than double in the AR other than 500-MW wind in this scenario. scenario, leading to an increase in total costs. Relatively minor differences in other categories The total cost in the AR scenario is roughly 7% in either scenario can be explained by the larger than in the CT scenario with the Excel fundamentally different approaches of the two

FIGURE 33 Figure 33 – Annual Average Prices and Reserve Margins with the Excel model Annual Average Prices and Reserve Margins with the Excel model

$60 CT 25% $60 AR 25%

$50 $50 20% 20%

$40 $40 15% 15% Margin Margin

$30 $30 ($2015/MWh) ($2015/MWh)

10% 10% Reserve $20 $20 Reserve Price Price 5% 5% $10 $10

$0 0% $0 0%

RM Price RM Price

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 33 Total Generation The Excel model yields very similar results to those from AURORAxmp, especially in the AR scenario. In CT scenario, the Excel model substitutes about 200 million MWh of coal with gas.

Figure 34 – Total Generation Output by Fuel Type from 2015 to 2030 with the Excel model 7000

6000 Wind 1,109 Wind Total Generation5000 1,684 The Excel model yields very similar results to those from AURORAxmp, especially in the AR scenario. In 4000 CT scenario, the Excel model substitutesNatural about Gas 200 million MWh of coal with gas. 2,972 Natural Gas 3000 Figure 34 – Total Generation Output by Fuel Type from 2015 to 20302,636 with the Excel model Million MWh Million FIGURE 34 20007000 Total Generation Output by Coal Coal 1,423 Fuel Type from 2015 to 2030 10006000 Wind 1,123 with the Excel model Nuclear1,109, 650 NuclearWind, 648 50000 1,684 Current Trends Aggressive Renewables 4000 Other HydroNaturalNuclear Gas Coal Natural Gas Wind Solar 2,972 Natural Gas 3000 2,636 Million MWh Million System Costs 2000 System costs with the Excel model are consistentCoal with those from AURORAxmpCoal (Figure 35). Total costs are almost the same1000 with the CT scenario1,423 and within 3% with the AR scenario.1,123 With both models, the shares of different cost categories areNuclear roughly, 650 the same in each scenarioNuclear. Fuel, 648 costs account for about 0 half of the total costs in the CT scenarioCurrent but Trends only 42% in the AR scenario.Aggressive Despite Renewables the fuel cost savings, capital costs more than double in the AR scenario, leading to an increase in total costs. Other Hydro Nuclear Coal Natural Gas Wind Solar Figure 35 – Total System Costs from 2015 to 2030 with the Excel model FIGURE 35 System Costs $200 Total System Costs from 2015 $180 to 2030 with the ExcelS modelystem costs with the Excel model are consistent with those from AURORAxmp (Figure 35). Total costs are almost the same$160 with the CT scenario and within 3% with the AR scenario.82.9 With both models, the shares of different$140 cost categories are roughly97.1 the same in each scenario. Fuel costs account for about half of the total costs$120 in the CT scenario but only 42% in the AR scenario. Despite the fuel cost savings, 20.5 capital costs more $100than double in the AR scenario, leading to an increase in total costs.

$80 Figure 35 – Total System24.0 Costs from 2015 to 2030 with the Excel model Billion $2015Billion 58.3 $200$60 $180$40 50.4 $20 $160 38.082.9 14.6 $0 $140 97.1 Current Trends Aggressive Renewables $120 20.5 $100 Base Capital Carrying Fixed O&M Variable O&M Fuel

$80 24.0 Billion $2015Billion 58.3 models. AURORAxmp calculates each$60 cost item economic at any load level from this case study. on an hourly basis across eight zones.$40 As such, There50.4 is one CC type dominating most of the individual plant characteristics, transmission baseload levels due to the relatively low fixed cost $20 38.0 52 constraints across zones, natural gas price and14.6 low fuel cost (low heat rate). One CT type is $0 basis differentials and seasonal variations, and Currentthe Trendscheapest technology at peakAggressive load Renewables levels, having other factors all impact the results. In contrast, the lowest fixed cost. When existing capacity is cost calculations for the Excel model are ex-Base Capital Carryingnot considered,Fixed O&M the systemVariable consists O&M of 49.2Fuel GW of post annual calculations based on aggregate New CC and 21.4 GW of New CT. When the forced assumptions such as annual average natural gas outage rates are considered, the New CC needed price applied to all gas generation across ERCOT. is 51.8 GW with a forced outage rate of 5% (49.2/ (1-0.05)), and the New CT needed is 22.5 GW. Screening Curve Method 52 Next, we consider the retirement of existing Among the six potential technologies, we chose capacity. A generation unit will retire when it is the cheapest three to plot in Figure 36. New coal more expensive to retrofit than to build a new power plants are not added because they are not unit of any technology. Existing units usually have

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 34 lower efficiency (and thus higher fuel cost), but their place in the least-cost solution from SCM the annual retrofit cost is usually cheaper than (right panel of Figure 37). As a result, a total of 3.3 The total cost in the AR scenario is roughly 7% larger than in the CT scenario with the Excel model, whereas the increase isthe only annualized 5% with AURORAxmp. capital cost Cost of theestimates new units. for each A category inGW the ofAR coal scenario capacity will be retired before 2030. are closer to each othergenerator (3–9%) across owner the would two models only butretrofit there isan a existingsignificant differenceHence, in base only capital 16.5 GW of existing coal capacity carrying cost in the CTunit scenario: if it is$14.6 cheaper billion than in the building Excel model a new versus unit. $18.7 billion inis AURORAxmp included for the following calculations. (28% difference). This Therefore,difference can we be needexplained to compare by the cost the of total the 4,700cost -MWcurve new gas builds with The remaining existing capacity and new AURORAxmp; the Excelof model the existing does not generation build any new with units the other new than generation 500-MW wind in this scenario. to check if existing generators are “qualified” to technologies are positioned on the horizontal Relatively minor differences in other categories in either scenario can be explained by the fundamentally remain in operation or should be retired. Note axis to minimize the overall system cost (shaded different approaches of the two models. AURORAxmp calculates each cost item on an hourly basis that their capital cost is already sunk, and only area in Figure 38). The existing technology across eight zones. As such, individual plant characteristics, transmission constraints acrosscurves zones, are added in the bottom denoted by natural gas price basisa differentials retrofit cost and is seasonal incurred. variations, Furthermore, and other in factors this all impact the results. In study, different coal power plants have different solid curves, and the new technology curves contrast, cost calculations for the Excel model are ex-post annual calculations based onremain aggregate the same denoted by dashed curves. assumptions such as annualretrofit average costs, natural so we gas need price to applied test them to all one gas generationby one. across ERCOT. When the existing units are added, all potential Screening Curve MethodIn most cases, the retrofit costs are low enough CTs are replaced by the existing CC2, CT1, and Among the six potentialsuch technologies, that the existing we chose coal the cheapestunit, with three retrofit to plot costsin Figure 36CT2;. New some coal power potential CC capacity is replaced by plants are not added becauseand lower they efficiency, are not economic is still at anycheaper load levelthan from all otherthis case study.existing There nuclear, is one coal, CC1, and CC2. Thus, there CC type dominating mostnew of units the baseload (left panel levels ofdue Figure to the 37).relatively However, low fixed cost anisd lowonly fuel new cost CC added to the existing system. (low heat rate). One CTsome type coalis the units cheapest with technology higher retrofit at peak loadcosts levels, lose having the lowest fixed cost. When existing capacity is not considered, the system consists of 49.2 GW of New CC andA high-level 21.4 GW of explanation about the existing New CT. When the forced outage rates are considered, the New CC needed is 51.8 GWcapacity with a forced positions is that they are ranked by their outage rate of 5% FIGURE 36 , and the New CT needed is 22.5 GW variable. fuel costs (VFCs) along with the new technologies (Table 17). The SCM dispatches (49SCM:.2/( ERCOT1 − 0. 052030)) CT Scenario without Existing Capacity (21.4/(1 − 0.1)) Figure 36 – SCM: ERCOT 2030 CT Scenario without Existing Capacity the technology with the lowest VFC first and then searches for the next cheapest technology. Hence, the cheaper technologies are dispatched as baseload generation and are allocated to lower load levels. Similarly, the most expensive technology in VFC is dispatched last.

Note that existing nuclear has lower operating costs than any new technology, so it is located at In most cases, the retrofit costs are low enough such that the lowestexisting load coal level; unit ,while with existingretrofit costCT1,s andCT2, and lower efficiency, is still cheaper than all other new units (leftCC2 panel units of are Figure more 37 expensive). However, to operate some coal than any units with higher retrofit costs lose their place in the least-new cost technology, solution from so theySCM are(right located panel at of the Figure highest 37). As a result, a total of 3.3 GW of coal capacity will be retiredload level. before Existing 2030. coal Hence, and only CC1 16.5 units GW have of VFCs Next, we consider37 the). Asretirement a result, of aexisting total ofcapacity 3.3 GW. A generationof coal capacity unit will will retire be when retired it is morebefore 2030. Hence, only 16.5 GW of existing coal capacity is included for the following calculations. expensive to retrofitexisting than to coal build capacity a new unit is includedof any technology for the. followinExisting unitsg calculations. usually have lower FIGURE 37 efficiency (and thus higher fuel cost), but the annual retrofit cost is usually cheaper than the annualized Figure 37 – SCM: The Retirement of a Coal Unit. Left: Not retired. Right: Retired. capital cost of the newSCM: units. The ARetirement generator ofFigure ownera Coal 37 Unit.would – SCM: Left: only NotThe retrofitretired. Retirement Right: an existing ofRetired. a Coal unit Unit. if it Left: is cheaper Not retired. than Right: Retired. building a new unit. Therefore, we need to compare the total cost curve of the existing generation with the new generation to check if existing generators are “qualified” to remain in operation or should be retired. Note that their capital cost is already sunk, and only a retrofit cost is incurred. Furthermore, in this study, different coal power plants have different retrofit costs, so we need to test them one by one.

53

The remaining existing capacity and new technologies are positioned on the horizontal axis to minimize The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 35 the overall system cost (shaded area in Figure 38). The existing technology curves are added in the bottom denoted by solid curves, and the new technology curves remain the same denoted by dashed curves. When the existing units are added, all potential CTs are replaced by the existing CC2, CT1, and CT2; some potential CC capacity is replaced by existing nuclear, coal, CC1, and CC2. Thus, there is only new CC added to the existing system.

Figure 38 – SCM: ERCOT 2030 CT Scenario

A high-level explanation about the existing capacity positions is that they are ranked by their variable fuel costs (VFCs) along with the new technologies (Table 17). The SCM dispatches the technology with the lowest VFC first and then searches for the next cheapest technology. Hence, the cheaper

54

technologies are dispatched as baseload generation and are allocated to lower load levels. Similarly, the most expensive technology in VFC is dispatched last.

In most cases, the retrofit costs are low enough such that the existing coal unit, with retrofit costs and Table 17 – VFC Rankings lower efficiency, is still cheaper than all other new units (left panel of FigureRank 37). However,1 some2 coal 3 4 4 5 6 7 units with higher retrofit costs lose their place in the least-cost solutionTech from SCMExst. (right Nuc panelNew of Figure CC Exst. coal Exst. CC1 New CT Exst. CC2 Exst. CT1 Exst. CT2 37). As a result, a total of 3.3 GW of coal capacity will be retired before 2030.VFC Hence,6.3 only 16.5 GW22.7 of 25.2 29.5 29.9 35.1 42.1 45.6 existing coal capacity is included for the following calculations.

Figure 37 – SCM: The Retirement of a Coal Unit. Left: Not retired.Note Right:that existingRetired. nuclear has lower operating costs than any new technology, so it is located at the lowest load level; while existing CT1, CT2, and CC2 units are more expensive to operate than any new technology, so they are located at the highest load level. Existing coal and CC1 units have VFCs between New CC and new CT, so their load level should have been between New CC and New CT (around 49.2 GW crossing point in Figure 38). However, the total capacity of existing CC2, CT1, and CT2 is more than the potential new CT, so the existing coal and CC1 are pushed to the left of the crossing point. The consideration of existing units also results in no new CT being built.

The values showing on the horizontal axis are after de-rating. In order to obtain the actual capacity MW or convert actual capacity to the values on the horizontal axis, we need to use forced outage rates (FOR). For example, when we consider 16.5 GW of existing coal, it can only reliably balance of load because it has a 10% FOR. So, the length of the projection of the red solid curve on the horizontal axis is 14.9 GW instead of 16.5 GW. Similarly, the calculated new CC16 from.5 × SCM is ( ) The remaining existing capacity and new technologies are positionedthe1 on− de the0-.rated1 horizontal= 14capacity.9 𝐺𝐺𝐺𝐺 axis of to 19.2 minimize GW. Its capacity should be converted to the actual capacity, which is the overall system cost (shaded area in Figure 38). The existing technology curves are added in the. bottom denoted by solid curves, and the new technology curves remain the same denoted by dashed curves. When the existing units are added, all potential CTs are replaced19The.2 only /by(1 the −difference 0existing.05) = CC2, between20.2 CT1, 𝐺𝐺𝐺𝐺 and the AR and the CT scenarios is that the wind and solar capacities are CT2; some potential CC capacity is replaced by existing nuclear, coal,increased, CC1, and CC2. and Thus,consequently there is only the net load is reduced. In this case, total thermal capacity is decreased new CC added to the existing system. from the Current Trends scenario, resulting in fewer new CC units needed (Figure 39). FIGURE 38 FIGURE 39 SCM: ERCOT 2030Figure CT Scenario 38 – SCM: ERCOT 2030 CT Scenario SCM: ERCOT 2030Figure AR Scenario39 – SCM: ERCOT 2030 AR Scenario

A high-level explanationbetween about New the existing CC and capacity new positionsCT, so theiris thatA loadthey summar are level rankedy of the by simulation theirthe redvariable solid results curve is provided on the in Tablehorizontal 18. Under axis the is CT 14.9 scenario, we have 3.1 GW fuel costs (VFCs) alongshould with havethe new been technologies between (Table New 17 CC). The and SCMmore New dispatches CC CT as compared the technologyGW to theinstead with AR scenario, of 16.5 which GW. has Similarly, 12 GW more the ofcalculated wind and 2.6 GW more of solar than the lowest VFC first(around and then 49.2searches GW for crossing the next cheapest point intechnology. Figurethe CT 38).Hence, scenario the. cheaper new CC from SCM is the de-rated capacity of 19.2 However, the total capacity of existing CC2, CT1, GW. Its capacity should be converted to the actual 55 capacity, which54 is 19.2/(1 - 0.05) = 20.2 GW. and CT2 is more than the potential new CT, so the existing coal and CC1 are pushed to the left of the crossing point. The consideration of existing The only difference between the AR and the CT units also results in no new CT being built. scenarios is that the wind and solar capacities are increased, and consequently the net load is The values showing on the horizontal axis are after reduced. In this case, total thermal capacity is de-rating. In order to obtain the actual capacity decreased from the Current Trends scenario, MW or convert actual capacity to the values on resulting in fewer new CC units needed (Figure 39). the horizontal axis, we need to use forced outage rates (FOR). For example, when we consider 16.5 A summary of the simulation results is provided GW of existing coal, it can only reliably balance in Table 18. Under the CT scenario, we have 16.5 × (1 - 0.1) = 14.9 GW of load because it has 3.1 GW more CC as compared to the AR a 10% FOR. So, the length of the projection of scenario, which has 12 GW more of wind and 2.6 GW more of solar than the CT scenario.

TABLE 17 VFC Rankings

Rank 1 2 3 4 4 5 6 7 Tech Exst. Nuc New CC Exst. coal Exst. CC1 New CT Exst. CC2 Exst. CT1 Exst. CT2 VFC 6.3 22.7 25.2 29.5 29.9 35.1 42.1 45.6

TABLE 18 SCM 2030 Capacities under the Two Scenarios (GW)

New Total Total Existing Existing Existing Existing Existing Existing Total CC Wind Solar Nuclear Coal CC1 CC2 CT1 CT2 Capacity

CT 20.2 20.3 0.9 5.1 16.5 2.4 32.2 2 3 102.6

AR 17.1 32.4 3.5 5.1 16.5 2.4 32.2 2 3 114.2

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 36 TABLE 19 Comparison of 2030 Capacities from the Three Models (GW)

Model Total Nuclear Total Coal Total NG Total Wind Total Solar Total Capacity*

AURORAxmp 5.1 18.9 60.9 20.3 0.9 106.8 CT SCM 5.1 16.5 59.8 20.3 0.9 102.6 Excel 5.1 18.9 56.7 20.8 0.9 103.3 AURORAxmp 5.1 13.3 62.6 32.4 3.5 117.6 AR SCM 5.1 16.5 56.7 32.4 3.5 114.2 Excel 5.1 18.9 56.7 32.4 3.1 117.1 * AURORAxmp and Excel totals include other generation such as hydro and biomass.

Comparison of Capacity does not consider wind and solar new builds, Expansion Results but this does not matter as AURORAxmp and Excel do not build much wind or solar beyond The SCM and Excel models treat the whole those hardwired. The most noticeable differences generation system as a single node, whereas occur for coal and gas capacities. AURORAxmp AURORAxmp has eight zones as depicted in yields more gas-fired capacity than the other Figure 1. However, in test runs with and without two models, especially in the AR scenario. The transmission constraints, we did not find that SCM gas capacity is only 1.1 GW less than that of transmission constraints in AURORAxmp caused AURORAxmp in the CT scenario, but Excel gas large deviations in capacity results. Accordingly, capacity is 4.2 GW less. Both SCM and Excel have we are reporting the AURORAxmp capacity the same gas capacity in the AR scenario, which is results from the transmission-constrained 5.9 GW less than that of AURORAxmp. The AR case as it is more realistic (Table 19). scenario results can be explained by AURORAxmp retiring 5.6 GW more coal capacity than the Excel All three models yield similar results. Total model and 3.2 GW more than the SCM. capacity estimates are within 4% for the CT scenario and 3% for the AR scenario. The SCM

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 37 4 | 2030 Hourly Dispatch Results

Total Generation in 2030 The AURORAxmp coal generation is 14–30% larger than the PLEXOS and Excel models. One Generation by fuel is highly consistent across all possible reason for these differences could be the three hourly dispatch models: AURORAxmp, Excel, costs included in merit order by different models. and PLEXOS (Figure 40). With all models, wind AURORAxmp and PLEXOS include start-up costs, and solar generation increase, nuclear generation which can be significant for fast-start units. In stays roughly the same, coal generation declines turn, this could lead to more coal-fired generation significantly, and gas generation falls less than 10%. being dispatched. These cost comparisons In the CT scenario, the PLEXOS and Excel are fruitful areas for further investigation. model results are very similar, but they display Price Duration Curve a wider discrepancy for gas and coal compared to AURORAxmp. Here, the PLEXOS and Excel Running hourly dispatch also allows us to compare models replace some coal generation with the price duration curves (PDCs) between the gas generation. For example, roughly 20,000 two scenarios across the models (Figure 41). The GWh of coal generation is replaced by about results are similar for most of the hours, but the the same amount of gas generation when PDCs of the two scenarios are much closer with compared to the AURORAxmp results. PLEXOS than with AURORAxmp, which has a much wider range between the highest and lowest With the AR scenario, both the PLEXOS and prices. Nevertheless, annual average prices are close ExcelWith models the AR generate scenario, more both fromthe PLEXOS gas plants and Excel modelsonce generate some extreme more from prices gas are plants taken than into account. thanAURORAxmp AURORAxmp (about (about 10,000 10,000 GWh GWh and 20,000 and GWh, respectively). This is at the expense of coal (9,000– There are significant differences at the extremes, 20,00018,000 GWh, GWh) respectively). and wind (4,000 This– 8,000is at the GWh). expense Differences in wind generation are relatively small across yielding different annual averages: $32–33/MWh in of thecoal models (9,000–18,000 (3–8%). TheGWh) PLE andXOS windand Excel (4,000– results for coal and gas differ by more than 10% from both scenarios with PLEXOS versus $49/MWh in 8,000AURORAxmp. GWh). Differences The AURORAxmp in wind coal generation generation is 14–30% larger than the PLEXOS and Excel models. are relatively small across the models (3–8%). the CT scenario and $40/MWh in the AR scenario One possible reason for these differences could be the costs included in merit order by different models. The PLEXOS and Excel results for coal and gas with AURORAxmp. The vertical axis in Figure 41 is AURORAxmp and PLEXOS include start-up costs, which can be significant for fast-start units. In turn, this differ by more than 10% from AURORAxmp. truncated at $150/MWh, but in the AURORAxmp could lead to more coal-fired generation being dispatched. These cost comparisons are fruitful areas for further investigation. FIGURE 40 Total GenerationFigure Output 40 – Total by Fuel Generation Type in 2030 Output - Comparison by Fuel Type of AURORAxmp, in 2030 - Comparison PLEXOS and of AURORAxmp, Excel Results PLEXOS and Excel Results 450,000 400,000 Wind Wind Wind 70,834 76,606 69,826 Wind Wind Wind 350,000 116,296 119,937 111,103 300,000 Natural Gas Natural Gas Natural Gas 250,000 184,643 212,107 210,165 Natural Gas Natural Gas Natural Gas 200,000 190,768 180,229 202,846 150,000 Coal 100,000 Coal Coal Coal

Total Generation (GWh) Generation Total Coal Coal 98,924 121,179 100,393 50,000 70,060 79,605 61,149 38,887 38,533 40,608 38,638 37,913 40,606 - CT (PLEXOS) CT CT (Excel) AR (PLEXOS) AR AR (Excel) (AURORAxmp) (AURORAxmp)

Other Hydro Nuclear Coal Natural Gas Wind Solar

Price Duration Curve TheRunning Full Cost hourlyof Electricity dispatch (FCe-) also allows us to compare the priceModel duration Documentation curves and Results (PDCs) for ERCOTbetween Scenarios, the July two 2017 | 38 scenarios across the models (Figure 41). The results are similar for most of the hours, but the PDCs of the two scenarios are much closer with PLEXOS than with AURORAxmp, which has a much wider range between the highest and lowest prices. Nevertheless, annual average prices are close once some extreme prices are taken into account.

There are significant differences at the extremes, yielding different annual averages: $32–33/MWh in both scenarios with PLEXOS versus $49/MWh in the CT scenario and $40/MWh in the AR scenario with AURORAxmp. The vertical axis in Figure 41 is truncated at $150/MWh, but in the AURORAxmp model, there are 22 hours in the CT scenario and 13 hours in the AR scenario with prices higher than $1,000/MWh. In contrast, the maximum price with PLEXOS is $68/MWh in the CT scenario and $88/MWh in the AR scenario.28 Annual average prices are much closer: $33 and $34 in the CT scenario; and $32 and $31 in the AR scenario for PLEXOS and AURORAxmp, respectively.

28 Note that the version of the PLEXOS model used in this study does not have scarcity pricing other than the value of lost load ($10,000/MWh). Given the high reserve margin with our 2030 load profile and generation capacity, the model finds a feasible solution without allowing unserved energy at VoLL.

57

Beyond the scarcity hours, AURORAxmp prices are higher than the PLEXOS prices for the first 1,000 hours (50% or larger difference for the first 280 hours in the CT scenario, and the first 176 hours for the AR scenario). These differences explain why the average prices are higher with AURORAxmp even after excluding the prices higher than $1,000/MWh.

On the other hand, prices are lower with AURORAxmp in 2,599 hours and 3,347 hours, under the CT and AR scenarios, respectively. However, these occur during low price periods (the tail of PDC curves in Figure 41) and, hence, do not influence the annual average price significantly. The larger number of lower priced hours in the AR scenario is expected given the much larger amount of wind and solar FIGUREcapacity. 41

Price Duration CurvesFigure for 2030 41 – - Price Comparison Duration of CurveAURORAxmps for 2030 and - PLEXOSComparison Results of AURORAxmp and PLEXOS Results $150

$125

$100

$75

$50 Price ($/MWh)Price $25

$0

CT (PLEXOS) CT(AURORAxmp) CT (Excel) AR (PLEXOS) AR (AURORAxmp) AR (Excel)

FIGUREPrices in 42 the Excel model reflect the marginal price of the highest dispatched technology to meet the Totaltotal System ERCOT Cost demand in 2030 (as compared to eight zones in other models). There are two exceptions: the highest 6% of the hours were approximatedFigure 42 – Total based System on Cost historical in 2030 prices (Figure 5), and the 18 hours with the lowest thermal$14 stress were assigned prices of $0/MWh. $12 At a high level, the Excel and AURORAxmp prices are close6.6 for the 6.5highest 3006.3 to 400 hours, especially in $10 7.4 the CT scenario. For the lowest7.8 2,000 hours or7.4 so, the Excel results are close to those from PLEXOS, $8 1.4 1.4 1.5 especially for the CT$6 scenario. Otherwise, Excel prices are higher than prices from both PLEXOS and 1.4 1.6 1.7 AURORAxmp: $10 $2015Billion –$415/MWh in many hours. Prices from 3.6the Excel model3.7 are 3.4basically the same across 3.5 the two scenarios except$2 for3.2 a period between3.2 hours 3,200 and 6,400. The Excel model prices also 2.7 2.7 2.7 conform to recent history$0 ,1.0 which may1.0 explain some1.0 of the differences. For example, under the AR scenario the renewable buildout could lead to more than 18 hours with marginal prices of $0/MWh as assumed, lowering the average price. System Costs in 2030

Total system costs depict Basethe Capital same Carrying distributionFixed of O&M capitalVariable and fuel O&M costsFuel across all three models used for the 8,760 runs (Figure 42) as the long-term results discussed earlier (Figure 31). Capital costs, as represented by the base capital carrying cost, are larger with the AR scenario, but fuel cost savings help model,toConclusions compensate there are. As22 and ahours result, Future in total the Work CTcosts scenario under theand AR scenarioBeyond are the only scarcity slightly hours, larger. AURORAxmp prices are 13 Thehours Full inCost the of AR Electricity scenario project with aims prices to provide higher a multifacetedhigher understanding than the PLEXOS of costs associated prices for with the first 1,000 thangenerating $1,000/MWh. and delivering In contrast, electricity the to maximum end-users as price well as costshours associated (50% or with larger fuel difference procurement for and the first 58280 with externalities. PLEXOS Asis $68/MWhpart of the project, in the anCT improved scenario version and of LCOEhours was in offered, the CT with scenario, regional and variability the first 176 hours for and some externality costs included. Other studies investigated costs of transmission and distribution $88/MWh in the AR scenario.25 Annual average the AR scenario). These differences explain why the investments, integration of distributed energy resources, and related topics. prices are much closer: $33 and $34 in the CT average prices are higher with AURORAxmp even scenario;In this report, and $32 using and the $31 ERCOT in thegrid ARas an scenario example, we simulatedafter different excluding capacity the expansionprices higher paths than to $1,000/MWh. forcompare PLEXOS costs and associated AURORAxmp, with different respectively. generation portfolios. These were measured by capital and operating costs, average electricity prices, and reserve marginsOn, among the other others hand,. These prices metrics are do lower not with form an exhaustive list but offer sufficient insight for the purposesAURORAxmp of this report. in 2,599We use hours three differentand 3,347 hours, 25 Note that the version of the PLEXOS model used in this study does not approacheshave scarcity pricing for long other-term than thecapacity value of expansion lost load ($10,000/MWh). analysis: a commercialunder the dispatch CT and software, AR scenarios, AURORAxmp respectively.; an ExcelGiven model;the high reserve and the margin screening with our curve2030 load method. profile andWe generation have also However,used PLEXOS, these another occur dispatch during software, low price periods AURORAxmp,capacity, the model and finds the a Excelfeasible model solution to without conduct allowing hourly unserved runs for(the 2030 tail under of PDC two scenarios curves inwith Figure distinctly 41) and, differentenergy at VoLL. generation portfolios.

The results are consistent in terms of overall capacity, but there are some differences in terms of how capacity is built or retired over time, the mix of the generation portfolio, average prices, and reserve Themargins Full Cost. This of Electricity conclusion (FCe-) is not surprising given the differences acrossModel theseDocumentation approaches and Results. However for ERCOT, weScenarios, July 2017 | 39 demonstrated that it is possible to obtain similar results once key assumptions are identified and key dispatch characteristics are captured. As such, all models can be useful for certain types of analyses.

Future work may include more detailed investigation of scenario results. For example, we only provided annual averages for prices and reserve margins across ERCOT. Looking at some of these prices at an hourly and/or zonal level could reveal some localized issues like the need for new transmission. Also, we would like to evaluate the sensitivity of results to key inputs such as the price of natural gas and capital cost trends for various generation technologies. More challenging, but equally important, would be the investigation of incorporating emerging technologies such as storage and distributed energy resources.

59 hence, do not influence the annual average price and AURORAxmp: $10–15/MWh in many hours. significantly. The larger number of lower priced Prices from the Excel model are basically the hours in the AR scenario is expected given the same across the two scenarios except for a period much larger amount of wind and solar capacity. between hours 3,200 and 6,400. The Excel model prices also conform to recent history, which may Prices in the Excel model reflect the marginal explain some of the differences. For example, under price of the highest dispatched technology to the AR scenario the renewable buildout could lead meet the total ERCOT demand (as compared to more than 18 hours with marginal prices of $0/ to eight zones in other models). There are two MWh as assumed, lowering the average price. exceptions: the highest 6% of the hours were approximated based on historical prices (Figure System Costs in 2030 5), and the 18 hours with the lowest thermal stress were assigned prices of $0/MWh. Total system costs depict the same distribution of capital and fuel costs across all three models At a high level, the Excel and AURORAxmp prices used for the 8,760 runs (Figure 42) as the long- are close for the highest 300 to 400 hours, especially term results discussed earlier (Figure 31). Capital in the CT scenario. For the lowest 2,000 hours or so, costs, as represented by the base capital carrying the Excel results are close to those from PLEXOS, cost, are larger with the AR scenario, but fuel cost especially for the CT scenario. Otherwise, Excel savings help to compensate. As a result, total costs prices are higher than prices from both PLEXOS under the AR scenario are only slightly larger.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 40 5 | CONCLUSIONS AND FUTURE WORK

The Full Cost of Electricity project aims to provide a The results are consistent in terms of overall multifaceted understanding of costs associated with capacity, but there are some differences in terms of generating and delivering electricity to end-users as how capacity is built or retired over time, the mix of well as costs associated with fuel procurement and the generation portfolio, average prices, and reserve externalities. As part of the project, an improved margins. This conclusion is not surprising given the version of LCOE was offered, with regional differences across these approaches. However, we variability and some externality costs included. demonstrated that it is possible to obtain similar Other studies investigated costs of transmission results once key assumptions are identified and key and distribution investments, integration of dispatch characteristics are captured. As such, all distributed energy resources, and related topics. models can be useful for certain types of analyses.

In this paper, using the ERCOT grid as an Future work may include more detailed example, we simulated different capacity expansion investigation of scenario results. For example, paths to compare costs associated with different we only provided annual averages for prices and generation portfolios. These were measured by reserve margins across ERCOT. Looking at some of capital and operating costs, average electricity these prices at an hourly and/or zonal level could prices, and reserve margins, among others. These reveal some localized issues like the need for new metrics do not form an exhaustive list but offer transmission. Also, we would like to evaluate the sufficient insight for the purposes of this paper. sensitivity of results to key inputs such as the price We use three different approaches for long- of natural gas and capital cost trends for various term capacity expansion analysis: a commercial generation technologies. More challenging, but dispatch software, AURORAxmp; an Excel equally important, would be the investigation model; and the screening curve method. We have of incorporating emerging technologies such as also used PLEXOS, another dispatch software, storage and distributed energy resources. AURORAxmp, and the Excel model to conduct hourly runs for 2030 under two scenarios with distinctly different generation portfolios.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 41 REFERENCES

AWS Truepower (2012, May 23). “Simulation of Wind ERCOT (2014, December 16). “Impacts of Environmental Generation Patterns for the ERCOT Service Area” Regulations in the ERCOT Region” [PDF document]. [PDF document]. Retrieved from http://www.ercot. Retrieved from http://www.ercot.com/content/news/ com/content/committees/other/lts/keydocs/2013/ presentations/2014/Impacts%20of%20Environmental%20 AWS_Truepower_ERCOT_Wind_Patterns_report.pdf Regulations%20in%20the%20ERCOT%20Region.pdf

Black & Veatch (2012, February). “Cost and ERCOT (2015a). “GIS Report February 2015” Performance Data for Power Generation Technologies” [Excel spreadsheet]. Retrieved from http:// [PDF document]. Retrieved from http://bv.com/ ercot.com/gridinfo/resource/2015 docs/reports-studies/nrel-cost-report.pdf ERCOT (2015b). “Wind shapes 1997–2014” [Zip DOE (2016). “Renewable Electricity Production Tax Credit archive with Excel spreadsheets]. Retrieved from (PTC).” Retrieved July 2016 from https://energy.gov/ http://ercot.com/gridinfo/resource/2015 savings/renewable-electricity-production-tax-credit-ptc ERCOT (2015c). “Capacity, Demand and Reserves DSIRE (2016). “Business Energy Investment Tax Report – December 2015” [Excel spreadsheet]. Retrieved Credit.” Retrieved July 2016 from http://programs. from http://ercot.com/gridinfo/resource/2015 dsireusa.org/system/program/detail/658 ERCOT (2015d, December 31). “2016 ERCOT System EIA (2013, April). “Updated Capital Cost Estimates for Planning Long-Term Hourly Peak Demand and Energy Utility Scale Electricity Generating Plants” [PDF document]. Forecast” [PDF document]. Retrieved from http://www. Retrieved from http://www.eia.gov/forecasts/capitalcost ercot.com/content/gridinfo/load/forecast/Docs/2016_Long- Term_Hourly_Peak_Demand_and_Energy_Forecast.pdf EIA (2015). “Form EIA-923 detailed data – Year 2014.” Retrieved from https://www.eia.gov/electricity/data/eia923/ ERCOT (2015e, December 15). “2016 LTSA Scenario Update” [PowerPoint presentation]. Retrieved December 2015 from EIA (2016a, February 8). “Form EIA-860 detailed http://www.ercot.com/calendar/2015/12/15/31834-RPG data – Year 2014.” Retrieved from https:// www.eia.gov/electricity/data/eia860/ ERCOT (2016a). “Capacity, Demand and Reserves Report – May 2016” [Excel spreadsheet]. Retrieved EIA (2016b). “Electric Power Monthly with Data for February from http://ercot.com/gridinfo/resource/2016 2016 – Table 6.5, Planned U.S. Electric Generating Unit Additions” [PDF document]. Retrieved from http://www. ERCOT (2016b). “GIS Report February 2016” eia.gov/electricity/monthly/current_year/april2016.pdf [Excel spreadsheet]. Retrieved from http:// ercot.com/gridinfo/resource/2016 EIA (2016c). “Electric Power Monthly with Data for February 2016 – Table 6.6, Planned U.S. Electric Generating Unit ERCOT (2016c, July 1). Section 3.14.1, Reliability Must Run. Retirements” [PDF document]. Retrieved from http://www. In “ERCOT Nodal Protocols” [PDF document]. Retrieved eia.gov/electricity/monthly/current_year/april2016.pdf from http://www.ercot.com/mktrules/nprotocols/lib

EIA (2016d). “Levelized Cost and Levelized Avoided ERCOT (2016d). “Weather” [Image file]. Retrieved July Cost of New Generation Resources in the Annual Energy 2016 from http://www.ercot.com/about/weather Outlook 2016” [PDF document]. Retrieved from https:// www.eia.gov/outlooks/aeo/pdf/electricity_generation.pdf ERCOT (2016e). “Hourly Load Data Archives” [Zip archive with Excel spreadsheets]. Retrieved July 2016 EIA (2016e). “Updated Capital Cost Estimates for Utility Scale from http://www.ercot.com/gridinfo/load/load_hist/ Electricity Generating Plants” [PDF document]. Retrieved from https://www.eia.gov/analysis/studies/powerplants/capitalcost/ Energy Exemplar Pty Ltd (2016). “PLEXOS.” Retrieved from https://energyexemplar.com/ ERCOT (2011, June 21). “Review of the Potential Impacts of Proposed Environmental Regulations on the EPA (2016). “GHG PSD Permits Issued by EPA.” Retrieved ERCOT System, Revision 1” [PDF document]. Retrieved March 2016 from https://yosemite.epa.gov/r6/apermit.nsf/AirP from http://www.ercot.com/news/presentations/2011/ ERCOT_Review_EPA_Planning_Final.pdf EPIS, LLC (2016). “AURORAxmp.” Retrieved from http://epis.com/aurora_xmp/

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 42 FERC (2011, April 28). “Order on Market-Based Rates, Rhodes, J., King, C., Gülen, G., Olmstead, S., Dyer, J., Terminating Section 206 Proceeding, and Denying Hebner, R., Beach, F., Edgar, T., & Webber, M. (2017). Request for Rehearing and Clarification” [PDF document]. “A geographically resolved method to estimate levelized Retrieved from http://www.ferc.gov/EventCalendar/ power plant costs with environmental externalities.” Energy Files/20110428175148-ER08-656-007.pdf Policy, 102, 491–99. doi:10.1016/j.enpol.2016.12.025

FERC (2016). “FERC Staff Reports & Papers – Energy SNL (2016). “Power Projects Database” Infrastructure Updates” [PDF documents]. Retrieved March Retrieved 2 March 2016. 2016 from http://www.ferc.gov/legal/staff-reports.asp Solar Energy Industries Association (SEIA) (2016). Garrison, J. (2014). “A Grid-Level Unit Commitment “Major Solar Projects List” [Excel spreadsheet]. Assessment of High Wind Penetration and Utilization of Retrieved March 2016 from http://www.seia.org/ Compressed Air Energy Storage in ERCOT” [PDF document]. research-resources/major-solar-projects-list Retrieved from http://hdl.handle.net/2152/28428 TCEQ (2015). “Greenhouse Gas Permitting.” Retrieved Hahn, W., DiLellio, J., & Dyer, J. (2016). “Market-calibrated March 2016 from https://www.tceq.texas.gov/permitting/ Forecasts for Natural Gas Prices” [PDF document]. White air/guidance/newsourcereview/ghg/ghg-permitting.html Paper UTEI/2016-07-1, 2016. Retrieved from http://energy. utexas.edu/the-full-cost-of-electricity-fce/fce-publications/ TCEQ (2016). “NSR Guidance for Turbines – Turbine List” [Excel spreadsheet]. Retrieved February 2016 from ICF International, Inc. (ICF) (2016). “U.S. DOE Combined https://www.tceq.texas.gov/permitting/air/guidance/ Heat and Power Database – Combined Heat and Power newsourcereview/turbines/nsr_fac_turb.html Installations in Texas.” Retrieved March 2016 from https://doe.icfwebservices.com/chpdb/state/TX Texas Comptroller of Public Accounts (TCPA) (2016). “Chapter 313 School Value Limitation Agreement Documents” Lazard (2014, September). “Lazard’s Levelized Cost [PDF documents]. Retrieved March 2016 from http:// of Energy Analysis—Version 8.0” [PDF document]. texasahead.org/tax_programs/chapter313/applicants/ Retrieved from https://www.lazard.com/media/1777/ levelized_cost_of_energy_-_version_80.pdf Tidball, R., Bluestein, J., Rodriguez, N., & Knoke, S. (2010). “Cost and Performance Assumptions for Namovicz, C. (2013, June 17). “Assessing the Economic Modeling Electricity Generation Technologies” Value of New Utility-Scale Renewable Generation Projects” [PDF document]. Subcontract Report, Contract No. [PDF document]. Retrieved July 2016 from https://www. DE-AC36-08GO28308. doi:10.2172/993653 eia.gov/conference/2013/pdf/presentations/namovicz.pdf Townsend, A., Webber, M. (2012). “An Integrated Analytical NREL (2013). “Financing, Overhead, and Profit: An In- Framework for Quantifying the LCOE of Waste-to-Energy Depth Discussion of Costs Associated with Third-Party Facilities for a Range of Greenhouse Gas Emissions Financing of Residential and Commercial Photovoltaic Policy and Technical Factors.” Waste Management, Systems” [PDF document]. Retrieved July 2016 from 32(7), 1366–77. doi:10.1016/j.wasman.2012.02.006 http://www.nrel.gov/docs/fy14osti/60401.pdf United States Congress (2015). “Consolidated NREL (2014). “2014 Cost of Wind Energy Review” Appropriations Act, 2016.” U.S. Government [PDF document]. Retrieved July 2016 from http:// Publishing Office, Washington, D.C. www.nrel.gov/docs/fy16osti/64281.pdf Washington State Department of Revenue (2016). NREL (2016). “PVWatts Calculator.” Retrieved “Utility Cost of Capital Studies” [PDF document]. from http://pvwatts.nrel.gov/ Retrieved July 2016 from http://dor.wa.gov/Content/ FindTaxesAndRates/PropertyTax/prop_UtilCapStudies.aspx Potomac Economics (2016, June). “2015 State of the Market Report for the ERCOT Wholesale Electricity Markets” [PDF Zhang, T., Baldick, R., & Deetjen, T. (2015). “Optimized document]. Retrieved from https://www.potomaceconomics. generation capacity expansion using a further improved com/uploads/ercot_documents/2015_ERCOT_State_of_ screening curve method.” Electric Power Systems the_Market_Report_-_FINAL_update_6.21_.16_.pdf Research, 124, 47–54. doi:10.1016/j.epsr.2015.02.017

PUCT (2015, December). “Electric Industry Reports Zhang, T., & Baldick, R. (2016). “Consideration of Existing – New Electric Generating Plants in Texas Since 1995” Capacity in Screening Curve Method.” IEEE Transactions on [Excel spreadsheet]. Retrieved from http://www.puc. Power Systems, PP(99), 1–1. doi:10.1109/TPWRS.2016.2633399 texas.gov/industry/Electric/reports/Default.aspx

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 43 Appendix A: ERCOT Plant Database for End-of-Year 201526

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Austin Area LFG Travis AEN 2007 Biogas IC 6.4 9,800 7.50 3.00 20 25 1 1 20.00 0 Dallas-Fort Worth Area LFG Dallas North 2015 Biogas IC 35.8 9,800 7.50 3.00 20 25 1 1 20.00 0 Houston Area LFG Harris Houston 2002 Biogas IC 24.5 9,800 7.50 3.00 20 25 1 1 20.00 0 San Antonio Area LFG Bexar CPS 2013 Biogas IC 26.8 9,800 7.50 3.00 20 25 1 1 20.00 0 Lufkin Biomass Angelina North 2012 Biomass ST 45 9,894 29.82 4.83 25 0.54 6 8 0.00 0 Nacogdoches Power Nacogdoches North 2012 Biomass ST 105 9,894 29.82 4.83 25 0.54 6 8 0.00 0 Big Brown 1 Freestone North 1971 Coal-Lig ST 606 10,744 43.56 6.33 48 0.264 12 24 42.00 10 Limestone 2 Limestone North 1986 Coal-Lig ST 858 9,578 43.56 6.33 48 0.264 12 24 42.00 10 Martin Lake 1 Rusk North 1977 Coal-Lig ST 800 11,129 43.56 6.33 48 0.264 12 24 42.00 10 Martin Lake 2 Rusk North 1978 Coal-Lig ST 805 11,063 43.56 6.33 48 0.264 12 24 42.00 10 Monticello U2 Titus North 1975 Coal-Lig ST 535 10,949 43.56 6.33 48 0.264 12 24 42.00 10 Oak Grove SES Unit 1 Robertson North 2010 Coal-Lig ST 840 9,344 43.56 6.33 48 0.264 12 24 42.00 10 Oak Grove SES Unit 2 Robertson North 2011 Coal-Lig ST 825 9,305 43.56 6.33 48 0.264 12 24 42.00 10 San Miguel 1 Atascosa South 1982 Coal-Lig ST 391 12,179 43.56 6.33 48 0.264 12 24 42.00 10 Sandow U4 Milam South 1981 Coal-Lig ST 570 9,323 43.56 6.33 48 0.264 12 24 42.00 10 Sandow U5 Milam South 2010 Coal-Lig ST 600 11,118 43.56 6.33 48 0.264 12 24 42.00 10 Twin Oaks 1 Robertson North 1990 Coal-Lig ST 156 10,887 43.56 6.33 48 0.264 12 24 42.00 10 Twin Oaks 2 Robertson North 1991 Coal-Lig ST 156 10,838 43.56 6.33 48 0.264 12 24 42.00 10 Big Brown 2 Freestone North 1972 Coal-Sub ST 602 10,684 43.56 6.33 48 0.264 12 24 29.00 10 Coleto Creek Goliad South 1980 Coal-Sub ST 660 10,163 43.56 6.33 48 0.264 12 24 29.00 10 Fayette Power Project 1 Fayette AEN 1979 Coal-Sub ST 604 10,692 43.56 6.33 48 0.264 12 24 29.00 10 Fayette Power Project 2 Fayette LCRA 1980 Coal-Sub ST 599 10,710 43.56 6.33 48 0.264 12 24 29.00 10 Fayette Power Project 3 Fayette LCRA 1988 Coal-Sub ST 437 10,673 43.56 6.33 48 0.264 12 24 29.00 10 Gibbons Creek 1 Grimes North 1983 Coal-Sub ST 470 9,990 43.56 6.33 48 0.264 12 24 29.00 10 J K Spruce 1 Bexar CPS 1992 Coal-Sub ST 560 10,822 43.56 6.33 48 0.264 12 24 29.00 10 J K Spruce 2 Bexar CPS 2010 Coal-Sub ST 775 10,800 43.56 6.33 48 0.264 12 24 29.00 10 J T Deely 1 Bexar CPS 1977 Coal-Sub ST 420 14,056 43.56 6.33 48 0.264 12 24 29.00 10 J T Deely 2 Bexar CPS 1978 Coal-Sub ST 420 14,093 43.56 6.33 48 0.264 12 24 29.00 10 Limestone 1 Limestone North 1985 Coal-Sub ST 831 9,643 43.56 6.33 48 0.264 12 24 29.00 10 Martin Lake 3 Rusk North 1979 Coal-Sub ST 805 11,077 43.56 6.33 48 0.264 12 24 29.00 10 Monticello U1 Titus North 1974 Coal-Sub ST 535 10,926 43.56 6.33 48 0.264 12 24 29.00 10 Monticello U3 Titus North 1978 Coal-Sub ST 795 10,936 43.56 6.33 48 0.264 12 24 29.00 10 Oklaunion 1 Wilbarger West 1986 Coal-Sub ST 650 10,634 43.56 6.33 48 0.264 12 24 29.00 10 Sandy Creek 1 McLennan North 2013 Coal-Sub ST 970 9,335 43.56 6.33 48 0.264 12 24 29.00 10 W A Parish 5 Fort Bend Houston 1977 Coal-Sub ST 659 10,413 43.56 6.33 48 0.264 12 24 29.00 10 W A Parish 6 Fort Bend Houston 1978 Coal-Sub ST 658 10,349 43.56 6.33 48 0.264 12 24 29.00 10 W A Parish 7 Fort Bend Houston 1980 Coal-Sub ST 577 10,354 43.56 6.33 48 0.264 12 24 29.00 10 W A Parish 8 Fort Bend Houston 1982 Coal-Sub ST 610 10,349 43.56 6.33 48 0.264 12 24 29.00 10 Hydro LCRA LZ Travis LCRA 1951 Hydro HYDRO 280 - 0.00 0.00 0 25 0 0 0.00 50 Hydro North LZ Bosque North 2014 Hydro HYDRO 51.6 - 0.00 0.00 0 25 0 0 0.00 50 Hydro RCEC LZ Grayson RCEC 1948 Hydro HYDRO 80 - 0.00 0.00 0 25 0 0 0.00 50 Hydro South LZ Starr South 2005 Hydro HYDRO 67.7 - 0.00 0.00 0 25 0 0 0.00 50 Hydro West LZ Val Verde West 1983 Hydro HYDRO 75.8 - 0.00 0.00 0 25 0 0 0.00 50 Arthur Von Rosenberg 1 Bexar South 2000 NG CCGT 458 8,017 25.28 4.73 25 0.42 6 14 34.63 0 B M Davis 3 Nueces South 2010 NG CCGT 633 8,318 25.28 4.73 25 0.42 6 14 34.63 0

26 FOM and VOM values represent average values used by the Excel model and SCM for the aggregate generation technologies. In AURORAxmp and PLEXOS, each unit has different values.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 44

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Bastrop Energy Center 1 Bastrop South 2002 NG CCGT 533 8,066 25.28 4.73 25 0.42 6 14 34.63 0 Bosque County CC 1 Bosque North 2009 NG CCGT 514 7,500 25.28 4.73 25 0.42 6 14 34.63 0 Bosque County CC 2 Bosque North 2001 NG CCGT 230.3 7,413 25.28 4.73 25 0.42 6 14 34.63 0 Brazos Valley 1 Fort Bend Houston 2003 NG CCGT 602 7,823 25.28 4.73 25 0.42 6 14 34.63 0 Cedar Bayou 4 Chambers Houston 2009 NG CCGT 504 7,041 25.28 4.73 25 0.42 6 14 34.63 0 Colorado Bend Energy Wharton Houston 2007 NG CCGT 233 7,416 25.28 4.73 25 0.42 6 14 34.63 0 Center 1 Colorado Bend Energy Wharton Houston 2008 NG CCGT 235 7,349 25.28 4.73 25 0.42 6 14 34.63 0 Center 2 Ennis Power Station 1 Ellis North 2002 NG CCGT 312 6,678 25.28 4.73 25 0.42 6 14 34.63 0 Ferguson Replacement Llano LCRA 2014 NG CCGT 509.8 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Forney Energy Center 1 Kaufman North 2003 NG CCGT 911 7,365 25.28 4.73 25 0.42 6 14 34.63 0 Forney Energy Center 2 Kaufman North 2003 NG CCGT 911 7,358 25.28 4.73 25 0.42 6 14 34.63 0 Freestone Energy Center 1 Freestone North 2002 NG CCGT 957.3 7,520 25.28 4.73 25 0.42 6 14 34.63 0 Frontera 1 Hidalgo South 1999 NG CCGT 524 6,840 25.28 4.73 25 0.42 6 14 34.63 0 Guadalupe Generating Guadalupe South 2000 NG CCGT 986 7,433 25.28 4.73 25 0.42 6 14 34.63 0 Station 1 Hays Energy Facility 1 Hays South 2002 NG CCGT 882 7,750 25.28 4.73 25 0.42 6 14 34.63 0 Hidalgo 1 Hidalgo South 2000 NG CCGT 458 7,167 25.28 4.73 25 0.42 6 14 34.63 0 Jack County Generation Jack North 2005 NG CCGT 595 8,752 25.28 4.73 25 0.42 6 14 34.63 0 Facility 1 Jack County Generation Jack North 2011 NG CCGT 595 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Facility 2 Johnson County Johnson North 1997 NG CCGT 269 8,368 25.28 4.73 25 0.42 6 14 34.63 0 Lamar Power Project 1 Lamar North 2000 NG CCGT 1,040 7,240 25.28 4.73 25 0.42 6 14 34.63 0 Lost Pines 1 Bastrop LCRA 2001 NG CCGT 528 7,651 25.28 4.73 25 0.42 6 14 34.63 0 Magic Valley Station Hidalgo South 2001 NG CCGT 670.2 7,326 25.28 4.73 25 0.42 6 14 34.63 0 Midlothian 1 Ellis North 2001 NG CCGT 940 8,074 25.28 4.73 25 0.42 6 14 34.63 0 Midlothian 2 Ellis North 2002 NG CCGT 504 7,639 25.28 4.73 25 0.42 6 14 34.63 0 Nueces Bay 8 Nueces South 2010 NG CCGT 633 8,231 25.28 4.73 25 0.42 6 14 34.63 0 Odessa Ector Generating Ector West 2001 NG CCGT 998.5 6,826 25.28 4.73 25 0.42 6 14 34.63 0 Station 1 Panda Sherman Grayson North 2014 NG CCGT 717 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Panda Temple 1 Bell North 2014 NG CCGT 702 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Panda Temple II CTG1 Bell North 2015 NG CCGT 191.2 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Panda Temple II CTG2 Bell North 2015 NG CCGT 191.2 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Panda Temple II STG Bell North 2015 NG CCGT 334.7 7,050 25.28 4.73 25 0.42 6 14 34.63 0 Paris Energy Center 1 Lamar North 1990 NG CCGT 239 7,395 25.28 4.73 25 0.42 6 14 34.63 0 Quail Run Energy 1 Ector West 2008 NG CCGT 488 6,500 25.28 4.73 25 0.42 6 14 34.63 0 Rio Nogales 1 Guadalupe South 2002 NG CCGT 785 8,050 25.28 4.73 25 0.42 6 14 34.63 0 Sam Rayburn 10 Victoria South 2003 NG CCGT 190 7,591 25.28 4.73 25 0.42 6 14 34.63 0 Sand Hill Energy Center 5a Travis AEN 2004 NG CCGT 295 8,069 25.28 4.73 25 0.42 6 14 34.63 0 Silas Ray 9 Cameron South 1996 NG CCGT 58 10,075 25.28 4.73 25 0.42 6 14 34.63 0 T H Wharton 3 Harris Houston 1974 NG CCGT 332 9,441 25.28 4.73 25 0.42 6 14 34.63 0 T H Wharton 4 Harris Houston 1974 NG CCGT 332 8,922 25.28 4.73 25 0.42 6 14 34.63 0 Tenaska Frontier Station Grimes Houston 2000 NG CCGT 880 7,307 25.28 4.73 25 0.42 6 14 34.63 0 Tenaska Gateway Station Rusk North 2001 NG CCGT 846 7,200 25.28 4.73 25 0.42 6 14 34.63 0 Tenaska Kiamichi Station 1 Fannin North 2003 NG CCGT 623 7,345 25.28 4.73 25 0.42 6 14 34.63 0 Tenaska Kiamichi Station 2 Fannin North 2003 NG CCGT 623 7,345 25.28 4.73 25 0.42 6 14 34.63 0 Victoria Power Station Victoria South 2009 NG CCGT 285 8,844 25.28 4.73 25 0.42 6 14 34.63 0 Wise County Power LP Wise North 2004 NG CCGT 665 7,596 25.28 4.73 25 0.42 6 14 34.63 0 Wolf Hollow Power Proj 1 Hood North 2002 NG CCGT 705 7,911 25.28 4.73 25 0.42 6 14 34.63 0 CCGT- Baytown Energy Center Chambers Houston 2002 NG 794.8 9,347 25.28 4.73 25 0.42 6 14 34.63 0 CHP

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 45

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] CCGT- C R Wing Cogen Plant Howard West 1988 NG 207.7 8,034 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Channel Energy Center Harris Houston 2001 NG 686.6 9,228 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Channelview Cogen Plant Harris Houston 2008 NG 861 6,614 25.28 4.73 25 0.42 6 14 34.63 0 CHP Clear Lake Cogenera- CCGT- Harris Houston 1985 NG 384.9 8,096 25.28 4.73 25 0.42 6 14 34.63 0 tion Ltd CHP Corpus Christi Energy CCGT- Nueces South 2002 NG 460.5 7,263 25.28 4.73 25 0.42 6 14 34.63 0 Center CHP CCGT- Deer Park Energy Center 1 Harris Houston 2014 NG 1,203 6,043 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Freeport Energy Center Brazoria Houston 2007 NG 63.4 8,017 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Green Power 2 Galveston Houston 2009 NG 287.2 12,251 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Gregory Power Facility San Patricio South 2000 NG 388.5 7,262 25.28 4.73 25 0.42 6 14 34.63 0 CHP Houston Chemical Com- CCGT- Harris Houston 2005 NG 52.6 9,500 25.28 4.73 25 0.42 6 14 34.63 0 plex Battleground CHP CCGT- Ingleside Cogeneration San Patricio South 1999 NG 478.4 8,000 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Optim Energy Altura Cogen Harris Houston 1995 NG 478.8 11,783 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Oyster Creek Unit VIII Brazoria Houston 1994 NG 404.5 10,753 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Pasadena Cogeneration Harris Houston 2000 NG 743.6 7,554 25.28 4.73 25 0.42 6 14 34.63 0 CHP CCGT- Texas City 1 Galveston Houston 2000 NG 458.5 14,485 25.28 4.73 25 0.42 6 14 34.63 0 CHP Wichita Falls Cogeneration CCGT- Wichita West 1987 NG 78 7,241 25.28 4.73 25 0.42 6 14 34.63 0 Plant CHP Greenville Powerlane IC1 Hunt North 2010 NG IC 8.4 9,833 19.49 13.41 20 25 1 1 19.37 0 Greenville Powerlane IC2 Hunt North 2010 NG IC 8.4 9,761 19.49 13.41 20 25 1 1 19.37 0 Greenville Powerlane IC3 Hunt North 2010 NG IC 8.4 9,828 19.49 13.41 20 25 1 1 19.37 0 Pearsall IC Engine Plant A Frio South 2012 NG IC 50.6 9,784 19.49 13.41 20 25 1 1 19.37 0 Pearsall IC Engine Plant B Frio South 2012 NG IC 50.6 9,789 19.49 13.41 20 25 1 1 19.37 0 Pearsall IC Engine Plant C Frio South 2012 NG IC 50.6 9,794 19.49 13.41 20 25 1 1 19.37 0 Pearsall IC Engine Plant D Frio South 2012 NG IC 50.6 9,788 19.49 13.41 20 25 1 1 19.37 0 Bryan Atkins 7 Brazos North 1973 NG OCGT 18 14,451 16.95 13.41 25 20 1 1 19.17 0 Dansby 2 Brazos North 2004 NG OCGT 45 10,584 16.95 13.41 24 44 1 1 18.98 0 Dansby 3 Brazos North 2010 NG OCGT 47 9,470 16.95 13.41 24 44 1 1 18.98 0 Decker Creek G1 Travis AEN 1989 NG OCGT 48 9,500 16.95 13.41 24 44 1 1 20.10 0 Decker Creek G2 Travis AEN 1989 NG OCGT 48 9,500 16.95 13.41 24 44 1 1 20.10 0 Decker Creek G3 Travis AEN 1989 NG OCGT 48 9,500 16.95 13.41 24 44 1 1 20.10 0 Decker Creek G4 Travis AEN 1989 NG OCGT 48 9,500 16.95 13.41 24 44 1 1 20.10 0 DeCordova 1 Hood North 1990 NG OCGT 71 9,850 16.95 13.41 25 14 1 1 36.96 0 DeCordova 2 Hood North 1990 NG OCGT 70 9,850 16.95 13.41 25 14 1 1 37.45 0 DeCordova 3 Hood North 1990 NG OCGT 69 9,850 16.95 13.41 25 14 1 1 37.96 0 DeCordova 4 Hood North 1990 NG OCGT 68 9,850 16.95 13.41 25 14 1 1 38.48 0 Ector County Energy Ector West 2015 NG OCGT 294 9,983 16.95 13.41 25 14 1 1 35.00 0 ExTex La Porte Power Harris Houston 2009 NG OCGT 38 9,250 16.95 13.41 24 44 1 1 18.75 0 Station AirPro 1 ExTex La Porte Power Harris Houston 2009 NG OCGT 38 9,250 16.95 13.41 24 44 1 1 18.75 0 Station AirPro 2 ExTex La Porte Power Harris Houston 2009 NG OCGT 38 9,250 16.95 13.41 24 44 1 1 18.75 0 Station AirPro 3 ExTex La Porte Power Harris Houston 2009 NG OCGT 38 9,250 16.95 13.41 24 44 1 1 18.75 0 Station AirPro 4 Greens Bayou 73 Harris Houston 1976 NG OCGT 46 10,268 16.95 13.41 25 21 1 1 20.71 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 46

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Greens Bayou 74 Harris Houston 1976 NG OCGT 46 10,268 16.95 13.41 25 21 1 1 20.71 0 Greens Bayou 81 Harris Houston 1976 NG OCGT 46 10,268 16.95 13.41 25 21 1 1 20.71 0 Greens Bayou 82 Harris Houston 1976 NG OCGT 58 10,268 16.95 13.41 25 21 1 1 20.45 0 Greens Bayou 83 Harris Houston 1976 NG OCGT 56 10,268 16.95 13.41 25 21 1 1 20.45 0 Greens Bayou 84 Harris Houston 1976 NG OCGT 46 10,268 16.95 13.41 25 21 1 1 21.16 0 Laredo Peaking 4 Webb South 2008 NG OCGT 90.1 10,056 16.95 13.41 50 21 1 1 19.37 0 Laredo Peaking 5 Webb South 2008 NG OCGT 87.3 10,056 16.95 13.41 50 21 1 1 20.00 0 Leon Creek Peaking 1 Bexar South 2004 NG OCGT 46 9,888 16.95 13.41 24 44 1 1 19.38 0 Leon Creek Peaking 2 Bexar South 2004 NG OCGT 46 9,888 16.95 13.41 24 44 1 1 19.38 0 Leon Creek Peaking 3 Bexar South 2004 NG OCGT 46 9,888 16.95 13.41 24 44 1 1 19.38 0 Leon Creek Peaking 4 Bexar South 2004 NG OCGT 46 9,888 16.95 13.41 24 44 1 1 19.38 0 Morgan Creek 1 Mitchell West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 36.11 0 Morgan Creek 2 Mitchell West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 36.11 0 Morgan Creek 3 Mitchell West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 36.11 0 Morgan Creek 4 Mitchell West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 36.11 0 Morgan Creek 5 Mitchell West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 36.11 0 Morgan Creek 6 Mitchell West 1988 NG OCGT 67 10,173 16.95 13.41 25 14 1 1 36.11 0 Permian Basin 1 Ward West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 32.98 0 Permian Basin 2 Ward West 1988 NG OCGT 65 10,173 16.95 13.41 25 14 1 1 34.68 0 Permian Basin 3 Ward West 1988 NG OCGT 68 10,173 16.95 13.41 25 14 1 1 34.62 0 Permian Basin 4 Ward West 1990 NG OCGT 69 10,078 16.95 13.41 25 14 1 1 34.14 0 Permian Basin 5 Ward West 1990 NG OCGT 70 10,078 16.95 13.41 25 14 1 1 34.36 0 R W Miller 4 Palo Pinto North 1994 NG OCGT 104 13,400 16.95 13.41 25 14 1 1 34.88 0 R W Miller 5 Palo Pinto North 1994 NG OCGT 104 12,873 16.95 13.41 25 14 1 1 34.88 0 Ray Olinger 4 Collin North 2001 NG OCGT 75 13,892 16.95 13.41 25 14 1 1 35.10 0 Sam Rayburn GT 1 Victoria South 1963 NG OCGT 11 14,572 16.95 13.41 25 21 1 1 20.35 0 Sam Rayburn GT 2 Victoria South 1963 NG OCGT 11 14,503 16.95 13.41 25 21 1 1 20.35 0 San Jacinto SES CTG 1 Harris Houston 1995 NG OCGT 81 13,488 16.95 13.41 25 14 4 4 37.04 0 San Jacinto SES CTG 2 Harris Houston 1995 NG OCGT 81 13,567 16.95 13.41 25 14 4 4 37.04 0 Sand Hill Energy Center Travis AEN 2001 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 18.78 0 GT 1 Sand Hill Energy Center Travis AEN 2001 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 18.37 0 GT 2 Sand Hill Energy Center Travis AEN 2001 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 19.18 0 GT 3 Sand Hill Energy Center Travis AEN 2001 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 20.00 0 GT 4 Sand Hill Energy Center Travis AEN 2010 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 18.78 0 GT 6 Sand Hill Energy Center Travis AEN 2010 NG OCGT 47 9,888 16.95 13.41 24 44 1 1 18.78 0 GT 7 Silas Ray 10 Cameron South 2004 NG OCGT 46 9,888 16.95 13.41 24 44 1 1 18.98 0 T H Wharton G 1 Harris Houston 1967 NG OCGT 13 10,268 16.95 13.41 25 20 1 1 19.37 0 T H Wharton GT 51 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.89 0 T H Wharton GT 52 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.80 0 T H Wharton GT 53 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.80 0 T H Wharton GT 54 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.80 0 T H Wharton GT 55 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.80 0 T H Wharton GT 56 Harris Houston 1975 NG OCGT 57 9,983 16.95 13.41 25 14 1 1 33.80 0 V H Braunig 5 Bexar South 2009 NG OCGT 48 9,894 16.95 13.41 24 44 1 1 18.78 0 V H Braunig 6 Bexar South 2009 NG OCGT 48 9,894 16.95 13.41 24 44 1 1 18.78 0 V H Braunig 7 Bexar South 2009 NG OCGT 48 9,894 16.95 13.41 24 44 1 1 18.78 0 V H Braunig 8 Bexar South 2009 NG OCGT 47 9,894 16.95 13.41 24 44 1 1 18.78 0 W A Parish Petra Nova Fort Bend Houston 2013 NG OCGT 74 9,983 16.95 13.41 25 14 1 1 35.00 0 W A Parish T1 Fort Bend Houston 1967 NG OCGT 13 14,467 16.95 13.41 25 20 1 1 19.17 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 47

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Winchester Power Park 1 Fayette LCRA 2009 NG OCGT 44 9,475 16.95 13.41 24 44 1 1 19.09 0 Winchester Power Park 2 Fayette LCRA 2009 NG OCGT 44 9,454 16.95 13.41 24 44 1 1 19.09 0 Winchester Power Park 3 Fayette LCRA 2009 NG OCGT 44 9,466 16.95 13.41 24 44 1 1 19.09 0 Winchester Power Park 4 Fayette LCRA 2009 NG OCGT 44 9,508 16.95 13.41 24 44 1 1 19.09 0 OCGT- Bayou Cogen Plant Harris Houston 1985 NG 165.6 9,927 16.95 13.41 25 14 1 1 33.89 0 CHP BP Chemicals Green OCGT- Calhoun South 1997 NG 8 11,500 16.95 13.41 25 14 1 1 33.88 0 Lake Plant CHP OCGT- Equistar Corpus Christi Nueces South 1989 NG 12 8,500 16.95 13.41 25 14 1 1 33.89 0 CHP ExxonMobil Baytown OCGT- Harris Houston 1989 NG 4.3 14,451 16.95 13.41 25 14 1 1 33.90 0 Refinery CHP OCGT- Sweeny Cogen Facility Brazoria Houston 2001 NG 430.7 9,000 16.95 13.41 25 14 1 1 33.89 0 CHP Texas Gulf Sulphur (New OCGT- Wharton Houston 1985 NG 58.9 14,553 16.95 13.41 25 14 1 1 33.90 0 Gulf) CHP OCGT- Victoria Texas Plant Victoria South 1987 NG 8.9 9,000 16.95 13.41 24 44 1 1 19.37 0 CHP B M Davis 1 Nueces South 1974 NG ST 330 10,500 19.49 15.43 38 0.54 2 2 47.94 0 Cedar Bayou 1 Chambers Houston 1970 NG ST 745 10,265 19.49 15.43 28 0.54 2 2 47.94 0 Cedar Bayou 2 Chambers Houston 1972 NG ST 749 10,271 19.49 15.43 28 0.54 2 2 47.94 0 Dansby 1 Brazos North 1978 NG ST 107 11,435 19.49 15.43 38 0.54 2 2 47.94 0 Decker Creek 1 Travis AEN 1971 NG ST 315 10,406 19.49 15.43 38 0.54 2 2 48.31 0 Decker Creek 2 Travis AEN 1978 NG ST 420 10,725 19.49 15.43 38 0.54 2 2 48.39 0 Graham 1 Young West 1960 NG ST 234 11,942 19.49 15.43 38 0.54 2 2 47.94 0 Graham 2 Young West 1969 NG ST 390 11,937 19.49 15.43 38 0.54 2 2 47.94 0 Greens Bayou 5 Harris Houston 1973 NG ST 371 13,460 19.49 15.43 38 0.54 2 2 50.00 0 Handley 3 Tarrant North 1963 NG ST 395 13,692 19.49 15.43 38 0.54 2 2 47.94 0 Handley 4 Tarrant North 1976 NG ST 435 13,692 19.49 15.43 38 0.54 2 2 48.42 0 Handley 5 Tarrant North 1977 NG ST 435 13,692 19.49 15.43 38 0.54 2 2 48.05 0 Lake Hubbard 1 Dallas North 1970 NG ST 392 12,162 19.49 15.43 38 0.54 2 2 47.94 0 Lake Hubbard 2 Dallas North 1973 NG ST 523 12,105 19.49 15.43 38 0.54 2 2 47.94 0 Mountain Creek 6 Dallas North 1956 NG ST 122 10,826 19.49 15.43 43 0.55 2 2 48.33 0 Mountain Creek 7 Dallas North 1958 NG ST 118 11,519 19.49 15.43 43 0.54 2 2 48.55 0 Mountain Creek 8 Dallas North 1967 NG ST 568 10,000 19.49 15.43 28 0.54 2 2 48.06 0 O W Sommers 1 Bexar South 1972 NG ST 420 12,058 19.49 15.43 38 0.54 2 2 46.79 0 O W Sommers 2 Bexar South 1974 NG ST 410 10,477 19.49 15.43 38 0.54 2 2 45.65 0 Pearsall 1 Frio South 1961 NG ST 19 14,500 19.49 15.43 43 0.52 2 2 47.94 0 Pearsall 2 Frio South 1961 NG ST 22 14,500 19.49 15.43 43 0.54 2 2 50.00 0 Pearsall 3 Frio South 1961 NG ST 20 14,500 19.49 15.43 43 0.54 2 2 50.00 0 Powerlane Plant 1 Hunt North 1966 NG ST 20 14,500 19.49 15.43 43 0.55 2 2 47.94 0 Powerlane Plant 2 Hunt North 1967 NG ST 26 14,500 19.49 15.43 43 0.54 2 2 48.86 0 Powerlane Plant 3 Hunt North 1978 NG ST 41 14,500 19.49 15.43 44 0.56 2 2 48.99 0 R W Miller 1 Palo Pinto North 1968 NG ST 75 10,947 19.49 15.43 43 0.53 2 2 47.94 0 R W Miller 2 Palo Pinto North 1972 NG ST 120 11,381 19.49 15.43 38 0.53 2 2 47.94 0 R W Miller 3 Palo Pinto North 1925 NG ST 208 10,335 19.49 15.43 38 0.54 2 2 47.94 0 Ray Olinger 1 Collin North 1967 NG ST 78 12,290 19.49 15.43 43 0.54 2 2 47.94 0 Ray Olinger 2 Collin North 1971 NG ST 107 11,351 19.49 15.43 38 0.53 2 2 47.94 0 Ray Olinger 3 Collin North 1975 NG ST 146 11,350 19.49 15.43 38 0.53 2 2 47.94 0 Sim Gideon 1 Bastrop LCRA 1965 NG ST 130 10,961 19.49 15.43 38 0.54 2 2 51.62 0 Sim Gideon 2 Bastrop LCRA 1968 NG ST 135 10,801 19.49 15.43 39 0.56 2 2 49.71 0 Sim Gideon 3 Bastrop LCRA 1972 NG ST 336 11,507 19.49 15.43 38 0.54 2 2 48.43 0 Spencer 4 Denton North 1966 NG ST 61 14,500 19.49 15.43 55 0.54 2 2 50.00 0 Spencer 5 Denton North 1973 NG ST 61 14,195 19.49 15.43 55 0.54 2 2 50.00 0 Stryker Creek 1 Cherokee North 1958 NG ST 167 13,850 19.49 15.43 39 0.55 2 2 49.08 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 48

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Stryker Creek 2 Cherokee North 1965 NG ST 502 13,908 19.49 15.43 38 0.54 2 2 50.00 0 Trinidad 6 Henderson North 1965 NG ST 235 13,510 19.49 15.43 38 0.54 2 2 47.94 0 V H Braunig 1 Bexar South 1966 NG ST 220 10,736 19.49 15.43 38 0.54 2 2 47.94 0 V H Braunig 2 Bexar South 1968 NG ST 230 10,278 19.49 15.43 38 0.53 2 2 47.94 0 V H Braunig 3 Bexar South 1970 NG ST 412 11,638 19.49 15.43 38 0.54 2 2 47.94 0 W A Parish 1 Fort Bend Houston 1958 NG ST 169 10,963 19.49 15.43 39 0.55 2 2 49.35 0 W A Parish 2 Fort Bend Houston 1958 NG ST 169 11,513 19.49 15.43 39 0.55 2 2 49.35 0 W A Parish 3 Fort Bend Houston 1961 NG ST 246 12,080 19.49 15.43 38 0.54 2 2 51.65 0 W A Parish 4 Fort Bend Houston 1968 NG ST 536 11,002 19.49 15.43 28 0.54 2 2 47.94 0 Acacia Solar Presidio West 2012 Solar PV 10 - 25.79 0.00 - - - - - 0 Barilla Solar (FS, Pecos) Pecos West 2014 Solar PV 22 - 25.79 0.00 - - - - - 0 Blue Wing 1 Solar Bexar South 2010 Solar PV 7.6 - 25.79 0.00 - - - - - 0 Blue Wing 2 Solar Bexar South 2010 Solar PV 7.3 - 25.79 0.00 - - - - - 0 Downie Ranch Solar (OCI Uvalde South 2015 Solar PV 95 - 25.79 0.00 - - - - - 0 Alamo 5) OCI Alamo 1 Solar Bexar South 2013 Solar PV 39.2 - 25.79 0.00 - - - - - 0 OCI Alamo 2-St. Hedwig Bexar South 2014 Solar PV 4.4 - 25.79 0.00 - - - - - 0 Solar OCI Alamo 3-Walzem Solar Bexar South 2014 Solar PV 5.5 - 25.79 0.00 - - - - - 0 OCI Alamo 4 Solar Kinney South 2014 Solar PV 37.6 - 25.79 0.00 - - - - - 0 (Bracketville) Renewable Energy Alterna- Denton North 2015 Solar PV 2 - 25.79 0.00 - - - - - 0 tives CCS1 SunEdison CPS3 Somerset Bexar South 2012 Solar PV 5.6 - 25.79 0.00 - - - - - 0 1 Solar SunEdison Rabel Road Bexar South 2012 Solar PV 9.9 - 25.79 0.00 - - - - - 0 Solar SunEdison Somerset Bexar South 2012 Solar PV 5 - 25.79 0.00 - - - - - 0 2 Solar SunEdison Valley Road Bexar South 2012 Solar PV 9.9 - 25.79 0.00 - - - - - 0 Solar Webberville Solar Travis AEN 2011 Solar PV 26.7 - 25.79 0.00 - - - - - 0 Notrees Battery Facility Winkler West 2012 Storage BATT 36 - 0.00 0.00 - - - - - 0 Comanche Peak U1 Somervell North 1990 Uranium ST 1,205 10,501 17.99 1.33 30 0.26 24 168 96.84 8 Comanche Peak U2 Somervell North 1993 Uranium ST 1,195 10,499 17.99 1.33 30 0.26 24 168 96.84 7 South Texas U1 Matagorda South 1988 Uranium ST 1,286 10,502 16.85 1.33 30 0.26 24 168 96.84 10 South Texas U2 Matagorda South 1989 Uranium ST 1,295 10,498 16.85 1.33 30 0.26 24 168 96.84 15 Anacacho Wind Kinney South 2012 Wind WT 99.8 - 41.18 0.00 - - - - - 0 Barton Chapel Wind Jack North 2007 Wind WT 120 - 41.18 0.00 - - - - - 0 Blue Summit Wind 5 Wilbarger West 2013 Wind WT 9 - 41.18 0.00 - - - - - 0 Blue Summit Wind 6 Wilbarger West 2013 Wind WT 126.4 - 41.18 0.00 - - - - - 0 Bobcat Bluff Wind Archer North 2012 Wind WT 150 - 41.18 0.00 - - - - - 0 Briscoe Wind Farm Briscoe West 2015 Wind WT 150 - 41.18 0.00 - - - - - 0 Buffalo Gap Wind Farm 1 Taylor West 2006 Wind WT 120.6 - 41.18 0.00 - - - - - 0 Buffalo Gap Wind Farm Taylor West 2007 Wind WT 115.5 - 41.18 0.00 - - - - - 0 2_1 Buffalo Gap Wind Farm Taylor West 2007 Wind WT 117 - 41.18 0.00 - - - - - 0 2_2 Buffalo Gap Wind Farm 3 Taylor West 2008 Wind WT 170.2 - 41.18 0.00 - - - - - 0 Bull Creek Wind Plant U1 Borden West 2009 Wind WT 88 - 41.18 0.00 - - - - - 0 Bull Creek Wind Plant U2 Borden West 2009 Wind WT 90 - 41.18 0.00 - - - - - 0 Callahan Wind Callahan West 2004 Wind WT 114 - 41.18 0.00 - - - - - 0 Camp Springs Wind 1 Scurry West 2007 Wind WT 130.5 - 41.18 0.00 - - - - - 0 Camp Springs Wind 2 Scurry West 2007 Wind WT 120 - 41.18 0.00 - - - - - 0 Capricorn Ridge Wind 1 Sterling West 2007 Wind WT 214.5 - 41.18 0.00 - - - - - 0 Capricorn Ridge Wind 2 Sterling West 2008 Wind WT 186 - 41.18 0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 49

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Capricorn Ridge Wind 3 Sterling West 2007 Wind WT 149.5 - 41.18 0.00 - - - - - 0 Capricorn Ridge Wind 4 Coke West 2008 Wind WT 112.5 - 41.18 0.00 - - - - - 0 Cedro Hill Wind 1 Webb South 2010 Wind WT 75 - 41.18 0.00 - - - - - 0 Cedro Hill Wind 2 Webb South 2010 Wind WT 75 - 41.18 0.00 - - - - - 0 Champion Wind Farm Nolan West 2008 Wind WT 126.5 - 41.18 0.00 - - - - - 0 Desert Sky Wind Farm 1 Pecos West 2002 Wind WT 84 - 41.18 0.00 - - - - - 0 Desert Sky Wind Farm 2 Pecos West 2002 Wind WT 76.5 - 41.18 0.00 - - - - - 0 Elbow Creek Wind Howard West 2008 Wind WT 118.7 - 41.18 0.00 - - - - - 0 Forest Creek Wind Farm Glasscock West 2007 Wind WT 124.2 - 41.18 0.00 - - - - - 0 Goat Wind Sterling West 2008 Wind WT 80 - 41.18 0.00 - - - - - 0 Goat Wind 2 Sterling West 2010 Wind WT 69.6 - 41.18 0.00 - - - - - 0 Goldthwaite Wind 1 Mills North 2014 Wind WT 148.6 - 41.18 0.00 - - - - - 0 Grandview 1 (Conway) Carson West 2014 Wind WT 107.4 - 41.18 0.00 - - - - - 0 GV1A Grandview 1 (Conway) Carson West 2014 Wind WT 103.8 - 41.18 0.00 - - - - - 0 GV1B Green Mountain Wind Scurry West 2003 Wind WT 99 - 41.18 0.00 - - - - - 0 (Brazos) U1 Green Mountain Wind Scurry West 2003 Wind WT 61 - 41.18 0.00 - - - - - 0 (Brazos) U2 Green Pastures Wind 1 Knox West 2015 Wind WT 150 - 41.18 0.00 - - - - - 0 Green Pastures Wind 2 Knox West 2015 Wind WT 150 - 41.18 0.00 - - - - - 0 Hackberry Wind Farm Shackelford West 2008 Wind WT 163.5 - 41.18 0.00 - - - - - 0 Hereford Wind G Deaf Smith West 2015 Wind WT 99.9 - 41.18 0.00 - - - - - 0 Hereford Wind V Deaf Smith West 2015 Wind WT 100 - 41.18 0.00 - - - - - 0 Horse Hollow Wind 1 Taylor West 2005 Wind WT 206.6 - 41.18 0.00 - - - - - 0 Horse Hollow Wind 2 Taylor West 2006 Wind WT 158 - 41.18 0.00 - - - - - 0 Horse Hollow Wind 3 Taylor West 2006 Wind WT 208 - 41.18 0.00 - - - - - 0 Horse Hollow Wind 4 Taylor West 2006 Wind WT 108 - 41.18 0.00 - - - - - 0 Inadale Wind Nolan West 2008 Wind WT 196.6 - 41.18 0.00 - - - - - 0 Indian Mesa Wind Farm Pecos West 2001 Wind WT 82.5 - 41.18 0.00 - - - - - 0 Javelina Wind Energy Zapata South 2015 Wind WT 250 - 41.18 0.00 - - - - - 0 Jumbo Road Wind 1 Deaf Smith West 2015 Wind WT 146.2 - 41.18 0.00 - - - - - 0 Jumbo Road Wind 2 Deaf Smith West 2015 Wind WT 153.6 - 41.18 0.00 - - - - - 0 Keechi Wind 138 KV Joplin Jack North 2014 Wind WT 110 - 41.18 0.00 - - - - - 0 King Mountain Wind NE Upton West 2001 Wind WT 79.3 - 41.18 0.00 - - - - - 0 King Mountain Wind NW Upton West 2001 Wind WT 79.3 - 41.18 0.00 - - - - - 0 King Mountain Wind SE Upton West 2001 Wind WT 40.3 - 41.18 0.00 - - - - - 0 King Mountain Wind SW Upton West 2001 Wind WT 79.3 - 41.18 0.00 - - - - - 0 Langford Wind Power Tom Green West 2009 Wind WT 155 - 41.18 0.00 - - - - - 0 Logan’s Gap Wind I U1 Comanche North 2015 Wind WT 103.8 - 41.18 0.00 - - - - - 0 Logan’s Gap Wind I U2 Comanche North 2015 Wind WT 106.3 - 41.18 0.00 - - - - - 0 Lone Star Wind 1 Shackelford North 2006 Wind WT 200 - 41.18 0.00 - - - - - 0 (Mesquite) Lone Star Wind 2 (Post Shackelford North 2007 Wind WT 100 - 41.18 0.00 - - - - - 0 Oak) U1 Lone Star Wind 2 (Post Shackelford North 2007 Wind WT 100 - 41.18 0.00 - - - - - 0 Oak) U2 Longhorn Wind North U1 Floyd West 2015 Wind WT 100 - 41.18 0.00 - - - - - 0 Longhorn Wind North U2 Floyd West 2015 Wind WT 100 - 41.18 0.00 - - - - - 0 Loraine Windpark I Mitchell West 2009 Wind WT 49.5 - 41.18 0.00 - - - - - 0 Loraine Windpark II Mitchell West 2009 Wind WT 51 - 41.18 0.00 - - - - - 0 Loraine Windpark III Mitchell West 2011 Wind WT 25.5 - 41.18 0.00 - - - - - 0 Loraine Windpark IV Mitchell West 2011 Wind WT 24 - 41.18 0.00 - - - - - 0 Los Vientos Wind III Starr South 2015 Wind WT 200 - 41.18 0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 50

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] McAdoo Wind Farm Dickens West 2008 Wind WT 150 - 41.18 0.00 - - - - - 0 Mesquite Creek Wind 1 Dawson West 2015 Wind WT 105.6 - 41.18 0.00 - - - - - 0 Mesquite Creek Wind 2 Dawson West 2015 Wind WT 105.6 - 41.18 0.00 - - - - - 0 Miami Wind G1 Gray West 2014 Wind WT 144.3 - 41.18 0.00 - - - - - 0 Miami Wind G2 Gray West 2014 Wind WT 144.3 - 41.18 0.00 - - - - - 0 Notrees Wind Farm 1 Winkler West 2009 Wind WT 92.6 - 41.18 0.00 - - - - - 0 Notrees Wind Farm 2 Winkler West 2009 Wind WT 60 - 41.18 0.00 - - - - - 0 Ocotillo Wind Farm Howard West 2008 Wind WT 58.8 - 41.18 0.00 - - - - - 0 Panhandle Wind 1 U1 Carson West 2014 Wind WT 109.2 - 41.18 0.00 - - - - - 0 Panhandle Wind 1 U2 Carson West 2014 Wind WT 109.2 - 41.18 0.00 - - - - - 0 Panhandle Wind 2 U1 Carson West 2014 Wind WT 94.2 - 41.18 0.00 - - - - - 0 Panhandle Wind 2 U2 Carson West 2014 Wind WT 96.6 - 41.18 0.00 - - - - - 0 Panther Creek 1 Howard West 2008 Wind WT 142.5 - 41.18 0.00 - - - - - 0 Panther Creek 2 Howard West 2008 Wind WT 115.5 - 41.18 0.00 - - - - - 0 Panther Creek 3 Howard West 2009 Wind WT 199.5 - 41.18 0.00 - - - - - 0 Pecos Wind (Woodward 1) Pecos West 2001 Wind WT 82.5 - 41.18 0.00 - - - - - 0 Pecos Wind (Woodward 2) Pecos West 2001 Wind WT 77.2 - 41.18 0.00 - - - - - 0 Pyron Wind Farm Scurry West 2008 Wind WT 249 - 41.18 0.00 - - - - - 0 Rattlesnake Den Wind Glasscock West 2015 Wind WT 104.3 - 41.18 0.00 - - - - - 0 1 G1 Rattlesnake Den Wind Glasscock West 2015 Wind WT 103 - 41.18 0.00 - - - - - 0 1 G2 Red Canyon Wind Borden West 2006 Wind WT 84 - 41.18 0.00 - - - - - 0 Roscoe Wind Farm Nolan West 2008 Wind WT 209 - 41.18 0.00 - - - - - 0 Route 66 Wind Carson West 2015 Wind WT 150 - 41.18 0.00 - - - - - 0 Sand Bluff Wind Farm Glasscock West 2008 Wind WT 90 - 41.18 0.00 - - - - - 0 Senate Wind Jack North 2012 Wind WT 150 - 41.18 0.00 - - - - - 0 Shannon Wind Clay West 2015 Wind WT 200 - 41.18 0.00 - - - - - 0 Sherbino 1 Wind Pecos West 2008 Wind WT 150 - 41.18 0.00 - - - - - 0 Sherbino 2 Wind Pecos West 2011 Wind WT 147.5 - 41.18 0.00 - - - - - 0 Silver Star Wind Eastland North 2008 Wind WT 60 - 41.18 0.00 - - - - - 0 Snyder Wind Farm Scurry West 2007 Wind WT 63 - 41.18 0.00 - - - - - 0 South Plains Wind 1 Floyd West 2015 Wind WT 102 - 41.18 0.00 - - - - - 0 South Plains Wind 2 Floyd West 2015 Wind WT 98 - 41.18 0.00 - - - - - 0 South Trent Wind Farm Nolan West 2008 Wind WT 98.2 - 41.18 0.00 - - - - - 0 Spinning Spur 3 Wind 1 Oldham West 2015 Wind WT 96 - 41.18 0.00 - - - - - 0 Spinning Spur 3 Wind 2 Oldham West 2015 Wind WT 98 - 41.18 0.00 - - - - - 0 Spinning Spur Wind Two Oldham West 2014 Wind WT 161 - 41.18 0.00 - - - - - 0 Stanton Wind Energy Martin West 2008 Wind WT 120 - 41.18 0.00 - - - - - 0 Stephens Ranch Wind 1 Borden West 2014 Wind WT 211.2 - 41.18 0.00 - - - - - 0 Stephens Ranch Wind 2 Borden West 2015 Wind WT 165 - 41.18 0.00 - - - - - 0 Sweetwater Wind 1 Nolan West 2003 Wind WT 36.6 - 41.18 0.00 - - - - - 0 Sweetwater Wind 2A Nolan West 2006 Wind WT 15.9 - 41.18 0.00 - - - - - 0 Sweetwater Wind 2B Nolan West 2004 Wind WT 97.5 - 41.18 0.00 - - - - - 0 Sweetwater Wind 3A Nolan West 2011 Wind WT 28.5 - 41.18 0.00 - - - - - 0 Sweetwater Wind 3B Nolan West 2011 Wind WT 100.5 - 41.18 0.00 - - - - - 0 Sweetwater Wind 4-4A Nolan West 2007 Wind WT 117.8 - 41.18 0.00 - - - - - 0 Sweetwater Wind 4-4B Nolan West 2007 Wind WT 103.7 - 41.18 0.00 - - - - - 0 Sweetwater Wind 4-5 Nolan West 2007 Wind WT 79.2 - 41.18 0.00 - - - - - 0 Texas Big Spring Wind A Howard West 1999 Wind WT 27.7 - 41.18 0.00 - - - - - 0 Texas Big Spring Wind B Howard West 1999 Wind WT 6.6 - 41.18 0.00 - - - - - 0 Trent Wind Farm Nolan West 2001 Wind WT 150 - 41.18 0.00 - - - - - 0 Trinity Hills Wind 1 Young North 2012 Wind WT 117.5 - 41.18 0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 51

Average Fixed Variable Start Heat O&M O&M Minimum Maximum Minimum Cost Forced Net Rate Charge Charge Stable Ramp Down Minimum [$/ Outage Prime Capacity [Btu/ [$/kW- [$/ Level Rate [%/ Time Up Time MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] [%] min] [hrs] [hrs] start] [%] Trinity Hills Wind 2 Young North 2012 Wind WT 107.5 - 41.18 0.00 - - - - - 0 TSTC West Texas Wind Nolan West 2008 Wind WT 2 - 41.18 0.00 - - - - - 0 Turkey Track Wind Energy Nolan West 2008 Wind WT 169.5 - 41.18 0.00 - - - - - 0 Center West Texas Wind Energy Upton West 1999 Wind WT 80.3 - 41.18 0.00 - - - - - 0 Whirlwind Energy Floyd West 2007 Wind WT 57 - 41.18 0.00 - - - - - 0 Whitetail Wind Energy Webb South 2012 Wind WT 91 - 41.18 0.00 - - - - - 0 Windthorst 2 Archer North 2014 Wind WT 67.6 - 41.18 0.00 - - - - - 0 WKN Mozart Wind Kent West 2012 Wind WT 30 - 41.18 0.00 - - - - - 0 Wolf Flats Wind (Wind Mgt) Hall West 2007 Wind WT 1 - 41.18 0.00 - - - - - 0 Wolf Ridge Wind Cooke North 2008 Wind WT 112.5 - 41.18 0.00 - - - - - 0 Cameron Wind Cameron South 2015 Wind-C WT 165 - 41.18 0.00 - - - - - 0 Gulf Wind I Kenedy South 2010 Wind-C WT 141.6 - 41.18 0.00 - - - - - 0 Gulf Wind II Kenedy South 2010 Wind-C WT 141.6 - 41.18 0.00 - - - - - 0 Harbor Wind Nueces South 2012 Wind-C WT 9 - 41.18 0.00 - - - - - 0 Los Vientos Wind I Willacy South 2013 Wind-C WT 200.1 - 41.18 0.00 - - - - - 0 Los Vientos Wind II Willacy South 2013 Wind-C WT 201.6 - 41.18 0.00 - - - - - 0 Magic Valley Wind Willacy South 2012 Wind-C WT 99.8 - 41.18 0.00 - - - - - 0 (Redfish) 1A Magic Valley Wind Willacy South 2012 Wind-C WT 103.5 - 41.18 0.00 - - - - - 0 (Redfish) 1B Papalote Creek Wind Farm San Patricio South 2009 Wind-C WT 179.9 - 41.18 0.00 - - - - - 0 Papalote Creek Wind San Patricio South 2010 Wind-C WT 200.1 - 41.18 0.00 - - - - - 0 Farm II Penascal Wind 1 Kenedy South 2009 Wind-C WT 160.8 - 41.18 0.00 - - - - - 0 Penascal Wind 2 Kenedy South 2009 Wind-C WT 141.6 - 41.18 0.00 - - - - - 0 Penascal Wind 3 Kenedy South 2011 Wind-C WT 100.8 - 41.18 0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 52 Appendix B: LCOE, LACE, and Net Value

Levelized cost of energy (LCOE) where CFn,t is the capacity factor for technology

n in period t, ΠFuelt is the cost of fuel in period TheLCOE is a method that quantifies the long- t, and is the heat rate of the technology. The term average cost for an electricity generation heat rate is the thermal energy needed to Appendix B: LCOE,technology LACE, andor facility, Net typically Value expressed in $/ generate a unit of electrical energy in plant n. kWh (Rhodes et al. 2017). The basic concept Levelized cost of energybehind (LCOE) LCOE is to combine the capital cost for an As shown in equation (4), the larger the capacity AppendixLCOE B: LCOE, LACE, and Net Value AppendixThe is a methodB: LCOE, thatelectric quantifiesLACE generation, and the long Net technology-term Value average n with cost its variablefor an electricity factor, generation the smaller technology the fixed and capital costs orLevelized facility, typically cost of expressed energyand (LCOE)fixed in $/kWh power (Rhodes generation et al. costs 2017) VOM( . Then,t basic + conceptrelative behind to LCOE each is unit to of energy, which lowers combineTheLevelized LCOE the is cost a capital method of energycost that FOMfor quantifies an(LCOE)n,t )electric in year the generation t. longThe- termcapital technology average costs are cost nconverted with for an its electricity variable LCOE and generation. fixedThis powerconcept—decreasing technology cost per unit into an annuity of equal sized payments over the of energy as a function of an increasing capacity generationorThe facility, LCOE istypically cost a methods ( expressed that+ quantifies in $/kWh) in the year (Rhodes long t. -Theterm et capital al. average 2017) costs. cost The are forbasic converted an concept electricity into behind generationan annuity LCOE is technologyof to equal generationAppendix period B so: LCOE,that the present LACE value, and of Net Valuefactor—is the key to understanding the competitive sizedcombineor facility, payments the typically capital over expressed cost the generationfor an in electric $/kWh period generation(Rhodes so that et technologytheal. 2017) present. The nvalue with basic ofits concept the variable annuity behind and is fixed the LCOE total power is toovernight 𝑉𝑉𝑉𝑉𝑉𝑉𝑛𝑛,the𝑡𝑡 Levelized 𝐹𝐹annuity𝐹𝐹𝐹𝐹𝑛𝑛,𝑡𝑡 is thecost total of overnightenergy (LCOE) capital cost. advantage of certain technologies over others. capitalgenerationcombine cost. the cost capitals ( cost for+ an electric) in yeargeneration t. The capitaltechnology costs n are with converted its variable into and an fixedannuity power of equal The LCOE is a method that quantifies the long-term average cost for an electricity generation technology Formally, we expresst the capital investment cost as: One of the desirable properties of LCOE is that sizedgeneration payments cost sover ( the𝑛𝑛, 𝑡𝑡generation+ 𝑛𝑛,𝑡𝑡) periodin year so . Thethat capital the present costs valueare converted of the annuity into an is annuitythe total of overnight equal Formally, we express𝑉𝑉𝑉𝑉𝑉𝑉 the(1) capital or 𝐹𝐹𝐹𝐹𝐹𝐹 facility, investment typically cost expressed as: in $/kWh (Rhodes et al.various 2017) technologies. The basic concept with different behind characteristicsLCOE is to capitalsized payments cost. over the generation period so that the present value of the annuity is the total overnight 𝑉𝑉𝑉𝑉𝑉𝑉𝑛𝑛,𝑡𝑡 combine 𝐹𝐹𝐹𝐹𝐹𝐹𝑛𝑛,𝑡𝑡 the capital cost for an electric generation technologycan be compared n with under its variable the same and cost fixed basis. power capital cost. For example, annual variable costs for wind Formally, we express the capitalgeneration investment costs (cost as: + ) in year t. The capital costs are converted into an annuity of equal 𝑇𝑇 and solar( 3are) zero since VOM is zero and these Formally, we express the capitalsized investmentpayments over cost the as: generation period so that the present value of the annuity is the total overnight 𝑉𝑉𝑉𝑉𝑉𝑉𝑛𝑛,𝑡𝑡 𝑅𝑅𝐹𝐹̅𝐹𝐹𝐹𝐹𝑛𝑛,𝑡𝑡 technologies do not require fuel. But, wind capital cost.𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 𝑛𝑛,𝑡𝑡 = ∑ 𝑡𝑡 and solar have relatively high capital costs 𝑡𝑡=𝑇𝑇1 (1 + 𝑖𝑖) (3) where is the equivalentwhere uniform R - is annualthe equivalent payment, uniform i is the annual discount rate and Twhen is the compared total number to a naturalof gas-fired power 𝑇𝑇 𝑅𝑅̅ (3) periods to service the debtpayment, Formally,(loan period) i is we the .express discountSolving𝑛𝑛,𝑡𝑡 thefor rate capital :and T investment is cost as:plant that does require fuel. LCOE allows for ̅ 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 = ∑ 𝑡𝑡 𝑅𝑅 the total number of periods𝑡𝑡=1 (to1 +service𝑅𝑅̅ 𝑖𝑖) the where is the equivalent uniform annual𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 𝑛𝑛payment,,𝑡𝑡 = ∑ i is the 𝑡𝑡discount rate and Tthese is the disparate total number features of of each technology to debt (loan period). Solving for𝑅𝑅̅ R -: periodswhere to is servicethe equivalent the debt uniform (loan period) annual. Solving payment,𝑡𝑡= for1 ( i1 is:+ the𝑖𝑖) discount rate and Tbe is fairlythe total compared number in of a single cost figure. ̅ (2) −1 𝑇𝑇 (3) 𝑅𝑅 𝑇𝑇 (4) periods to service the debt (loan period). Solving for : While𝑅𝑅̅ LCOE is a convenient measure of the overall 𝑅𝑅̅ 𝑅𝑅̅ 1 𝑛𝑛,𝑡𝑡 𝑡𝑡 𝑛𝑛,𝑡𝑡 𝑡𝑡 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 = ∑cost of different generating technologies, one of its 𝑅𝑅̅ = 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 × [∑ ] 𝑡𝑡=1 (1 + 𝑖𝑖) where is the equivalent𝑡𝑡=𝑇𝑇1 𝑅𝑅( ̅uniform1 + 𝑖𝑖) −annual1 payment, i is the discount rate and T is the total number of ⏟ biggest limitations(4) is that it does not consider the 𝐶𝐶𝐶𝐶𝐶𝐶 The factor on the right side periodsof the capital to service investment the debt𝑇𝑇 cost (loan1 in period)equation−1 . Solving (4) is the forrevenue capital : recoveryor(4) value of factor a technology, which, among 𝑅𝑅̅ =𝑅𝑅̅ 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝑛𝑛,𝑡𝑡 × [∑ 𝑡𝑡] (CRF) and can be simplified as follows. 𝑡𝑡=1 (1 +1 𝑖𝑖) other factors, is dependent on grid topography, 𝑅𝑅̅ = 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝑛𝑛,𝑡𝑡 × [⏟∑ 𝑡𝑡 ] 𝑅𝑅̅ The factor on the right sideThe of factor the capital on the investment right side𝑡𝑡=1 (ofcost1 𝐶𝐶𝐶𝐶𝐶𝐶the+ in𝑖𝑖capital) equation investment (4) is the load capital profiles, recovery generation factor portfolio, and fuel prices cost in equation (2) is the⏟ capital recovery 𝑇𝑇 at various hubs−1 in any given system. As a result, (TheCRF factor) and can on thebe simplified right side asof thefollows. capital investment cost𝐶𝐶𝐶𝐶𝐶𝐶 in equation (4) is the capital recovery factor (4) factor (CRF) 𝑇𝑇and can be simplified−1 as follows. an intermittent1 (5) resource like wind may have ̅ 𝑇𝑇 𝑛𝑛,𝑡𝑡 𝑡𝑡 (CRF) and can be simplified(3) as follows. 𝑅𝑅 = 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 × [∑the lowest] LCOE of all available technologies, 1 𝑖𝑖(1 + 𝑖𝑖) 𝑡𝑡=1 (1 + 𝑖𝑖) 𝑡𝑡 𝑇𝑇 ⏟ 𝐶𝐶𝐶𝐶𝐶𝐶 = [∑ ] = but𝐶𝐶𝐶𝐶𝐶𝐶 if this technology can be only dispatched The factor𝑡𝑡=𝑇𝑇 on1 ( the1 + right𝑖𝑖) − side1 of(1 the+ 𝑖𝑖 capital) − 1 investment cost in( 5equation) (4) is the capital recovery factor The annual LCOE can be calculated as (Tidball et al. 2010): 𝑇𝑇 at night in a system, when the wholesale price (CRF) and can𝑇𝑇 be1 simplified−1 as 𝑖𝑖follows.(1 + 𝑖𝑖) of electricity(5) is low, the technology may not be 𝐶𝐶𝐶𝐶𝐶𝐶 = [∑ 𝑡𝑡] = 𝑇𝑇 𝑇𝑇 𝑡𝑡=1 1 𝑖𝑖(1 + 𝑖𝑖) viable, even though it has the cheapest average The annual LCOE can be calculated as (Tidball(1 + 𝑖𝑖et)𝑡𝑡 al. 2010)(1: + 𝑖𝑖)𝑇𝑇 − 1 𝐶𝐶𝐶𝐶𝐶𝐶 = [∑ ] = cost per unit at(6 the) time that it is dispatched. The annual LCOE canΠ 𝐶𝐶𝐶𝐶𝐶𝐶be calculated𝑛𝑛,𝑡𝑡 × 𝐶𝐶𝐶𝐶𝐶𝐶 as+𝑡𝑡= (Tidball𝐹𝐹1𝐹𝐹𝐹𝐹(1 +𝑛𝑛,𝑡𝑡 𝑖𝑖et) al. 2010)(1: + 𝑖𝑖) − 1 $ 𝑛𝑛,𝑡𝑡 The annualLCOE can be calculated𝑛𝑛,𝑡𝑡 𝑛𝑛 𝑇𝑇 𝑡𝑡 −1 (5) 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 = 𝑛𝑛,𝑡𝑡 + ⏟𝑉𝑉𝑉𝑉𝑉𝑉 + 𝐻𝐻 𝐻𝐻 × Π 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 = 𝑇𝑇 ⏟ 8, 760 × 𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑘𝑘𝑘𝑘ℎ 𝐴𝐴𝐴𝐴𝐴𝐴as𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 (Tidball 𝐹𝐹𝐹𝐹𝐹𝐹 et𝐹𝐹𝐹𝐹 al.𝐶𝐶𝐶𝐶𝐶𝐶 2010):𝐶𝐶𝐶𝐶 1 𝑖𝑖(1 + 𝑖𝑖)(6) where is the capacityΠ𝐶𝐶𝐶𝐶𝐶𝐶𝑛𝑛, 𝑡𝑡factor× 𝐶𝐶𝐶𝐶𝐶𝐶 for+ technology𝐹𝐹𝐹𝐹𝐹𝐹𝑛𝑛,𝑡𝑡 n in period𝐶𝐶𝐶𝐶𝐶𝐶 =t, [∑ is the𝑡𝑡] cost$= of fuel in𝑇𝑇 period t, and 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑛𝑛,𝑡𝑡 = The annual LCOE can+ be𝑉𝑉𝑉𝑉𝑉𝑉 calculated𝑛𝑛,𝑡𝑡 + 𝐻𝐻𝐻𝐻 as𝑛𝑛𝑡𝑡= ×(Tidball1Π(𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹1 + 𝑖𝑖et𝑡𝑡)= al. 2010)(1: + 𝑖𝑖) −(6)1 is the heat rateΠ of𝐶𝐶𝐶𝐶𝐶𝐶 the(4)𝑛𝑛, 𝑡𝑡technology.× 𝐶𝐶𝐶𝐶𝐶𝐶 +𝑛𝑛𝐹𝐹,𝑡𝑡 The𝐹𝐹𝐹𝐹 𝑛𝑛heat,𝑡𝑡 ⏟rate is the thermal energy needed$ to generate a unit of 𝐶𝐶𝐶𝐶𝑛𝑛,𝑡𝑡 𝑛𝑛,𝑡𝑡 ⏟ 8, 760 × 𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝑛𝑛,𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝑛𝑛𝛱𝛱𝑉𝑉𝑉𝑉𝑉𝑉𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱 𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡𝐶𝐶𝐶𝐶 𝑡𝑡 𝑘𝑘𝑘𝑘ℎ whereelectrical 𝐿𝐿 𝐿𝐿𝐿𝐿energy 𝐿𝐿is the =in capacity plant𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴. factor 𝐹𝐹𝐹𝐹 for𝐹𝐹𝐹𝐹𝐹𝐹 𝑛𝑛technology 𝐶𝐶𝐶𝐶𝐶𝐶,𝑡𝑡 𝐶𝐶𝐶𝐶 + ⏟𝑉𝑉𝑉𝑉𝑉𝑉n in period + 𝐻𝐻 t 𝐻𝐻, × Π 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 is the= cost of fuel in period t, and 𝐻𝐻𝐻𝐻𝑛𝑛 ⏟ 8, 760 × 𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑘𝑘𝑘𝑘ℎ 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 n t, t, where is the𝑛𝑛 ,heat𝑡𝑡 is the rate capacity of the technology.factor for technology The heat ratein periodis the thermal energy𝑡𝑡 is the needed cost of to fuel generate in period a unitand of (6) As shown𝐶𝐶𝐶𝐶 in equation (6), 𝑛𝑛 the larger the capacityΠ𝐶𝐶𝐶𝐶𝐶𝐶 factor,𝑛𝑛,𝑡𝑡 ×the𝐶𝐶𝐶𝐶𝐶𝐶 smaller 𝛱𝛱+𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝐹𝐹𝐹𝐹𝐹𝐹 the𝑛𝑛 fixed,𝑡𝑡 and capital costs relative to $ electrical is the 𝑛𝑛energy ,heat𝑡𝑡 rate in plant of the. technology. The𝑛𝑛,𝑡𝑡 heat rate is the thermal energy𝑡𝑡 needed𝑛𝑛,𝑡𝑡 to generate𝑛𝑛 a unit𝑡𝑡 of 𝐻𝐻each𝐻𝐻𝑛𝑛 unit𝐶𝐶𝐶𝐶 of energy, which lowers LCOE𝐿𝐿𝐿𝐿𝐿𝐿.𝐿𝐿 This= concept—decreasing 𝛱𝛱𝑛𝑛𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱,𝑡𝑡 cost per+ unit⏟𝑉𝑉𝑉𝑉𝑉𝑉 of energy + 𝐻𝐻 𝐻𝐻 as × a Πfunction 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 = electrical energy in plant . ⏟ 8, 760 × 𝐶𝐶𝐶𝐶 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑘𝑘𝑘𝑘ℎ 𝐻𝐻ofAs𝐻𝐻 anshown𝑛𝑛 increasing in equation capacity (6), factor𝑛𝑛 thewhere la—rgeris the the key iscapacity the to capacityunderstanding factor,𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 factor the smaller𝐹𝐹𝐹𝐹 thefor𝐹𝐹𝐹𝐹𝐹𝐹 technologycompetitive 𝐶𝐶𝐶𝐶𝐶𝐶 the𝐶𝐶𝐶𝐶 fixed n andadvantage in period capital t of,costs certain relative is the to cost of fuel in period t, and technologieseachAs shown unit ofin energy,equation over others. which (6), 𝑛𝑛 the lowers larger is LCOEthe the heat. capacityThis rate concept of factor, the— technology.decreasing the smaller costThe the heatperfixed unit rate and of is capital energythe thermal costs as a functionrelative energy to needed to generate a unit of 𝐶𝐶𝐶𝐶𝑛𝑛,𝑡𝑡 𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝛱𝑡𝑡 ofeach an unitincreasing of energy, capacity which factor lowerselectrical—is LCOE the energy key. This to in concept understanding plant —. decreasing the competitive cost per unit advantage of energy of as certain a function One of the desirable propertiesThe𝐻𝐻 Full𝐻𝐻𝑛𝑛 Costof LCOE of Electricity is that (FCe-)various technologies with different Modelcharacteristics Documentation and can Results be for ERCOT Scenarios, July 2017 | 53 technologiesof an increasing over capacity others. factor —is the key to understanding the competitive advantage of certain compared under the same costAs shown basis. inFor equation example, (6 ),annual 𝑛𝑛 the la variablerger the costscapacity for factor,wind and the solar smaller are thezero fixed and capital costs relative to technologies over others. One of the desirable propertieseach of unit LCOE of energy,is that various which lowers technologies LCOE. Thiswith concept different— characteristicsdecreasing cost can per be unit of energy as a function One of the desirable propertiesof an of increasing LCOE is that capacity various factor technologies—is the key with to differentunderstanding characteristics the competitive can be advantage of certain compared under the same cost basis. For example, annual variable costs for wind and solar are zero 72 compared under the same costtechnologies basis. For over example, others. annual variable costs for wind and solar are zero One of the desirable properties of LCOE is that various technologies with different72 characteristics can be compared under the same cost basis. For example, annual variable costs for72 wind and solar are zero

72 since VOM is zero and these technologies do not require fuel. But, wind and solar have relatively high would have a LACE of $0/MWh, because it does capital costs when comparedLevelized to a natural avoided gas- firedcost power of energy plant that(LACE) does require fuel. LCOE allows for not sell any units of electricity, i.e. AAGn,t = 0 t , these disparate featuresIn of recent each technologyyears, the U.S. to beEnergy fairly Information compared in a single cost figure. since VOM is zero and theseAdministration technologies (EIA) do not has requiredeveloped fuel the. But, wind and solarIn order have to relatively be dispatched, high a conventional power capitalWhile LCOE costs is when a convenient comparedlevelized measure to a avoided natural of the costgas overall- offired electricity costpower of differentplant(LACE that) generating does requireplant technologies, fuel. must LCOE offer allowsone to ofgenerate forits electricity at a price thesebiggest disparate limitations features is thatas of it a each doescomplement technologynot consider to LCOE to the be . revenue faRatherirly compared than or value costs, in of a a single technology costthat figure.is, which,low enough among to otherenter into the market; less factors, is dependent onLACE grid topography, estimates the load revenue profiles, that generation a power plant portfolio, andefficient fuel prices plants at arevarious dispatched only at periods of higher demand when power prices are high hubsWhile in LCOE any givenis a convenient systemcreates. As measure a result,per each of an the unit intermittent overall of electricity cost resource of (Namovicz different like windgenerating may have technologies, the lowest one LCOE of ofits all 2013). LACE can be calculated as the weighted enough to cover their higher costs. This highlights availablebiggest limitations technologies, is that but it ifdoes this nottechnology consider can the be revenue only dispatched or value of at a night technology in a system, which,, when among the other average revenue that a certain technology an important feature of LACE: for conventional factors, is dependent on grid topography, load profiles, generation portfolio, and fuel prices at various wholesale price of electricitywould is provide low, the per technology unit of electricity, may not inbe a viable, even thoughpower it plants, has the LACE cheapest does not only depend on each averagehubs in any cost given per unit system at particularthe. As time a result, that period, anit is intermittent indispatched. $/kWh like resource LCOE. like wind may haveparticular the lowest technology, LCOE ofbut all also on the market as available technologies, but if this technology can be only dispatched at night ina awhole. system For, when example, the a certain technology in a wholesaleLevelized price avoided of electricity costOne of interpretationis energy low, the (LACE) technology of the LACE may of not a power be viable, plant even is thoughmarket it withhas thehigh cheapest prices would have a much higher that it is a measure of how much it would cost the LACE than the same technology in a low-price averageIn recent cost years per, the unit U.S. at theEnergy time Information that it is dispatched. Administration (EIA) has developed the levelized avoided grid to generate the additional electricity that would market. Therefore,LACE is not an appropriate LACE LCOE LACE cost of electricity ( )be as required a complement if that powerto plant. Rather were than not availablecosts, estimatesway to compare the revenue technologies that a across markets. powerLevelized plant avoided creates per cost (EIAeach of 2016d). unitenergy of electricityThe (LACE) word (Na“avoided”movicz is 2013) included. LACE in can the be calculated as the weighted averageIn recent revenue years, the that U.S. a certain LACEEnergy acronym technology Information to reflect would Administration theprovide cost persavings (EIA) unit hasassociated of electricity,developed Renewable inthe a levelizedparticular plants avoided period have a, marginalin cost of $0/MWh, and so they are always dispatched when available. $/kWhcost of electricitylike LCOE. (LACE)with as a avoidingcomplement the less to LCOE efficient. Rather generation than costs, that LACE estimates the revenue that a power plant creates per iseach replaced unit of by electricity the analyzed (Na technology.movicz 2013). LACE can be Therefore,calculated astheir theLACE weighted only depends on the market price and the availability of the generator that averageOne interpretation revenue that of athe certain LACE technologyof a power plantwould is provide that it is per a measure unit of electricity, of how much in a itparticular would cost period the , in The first concept to understand demandbefore calculating when power determines prices are thehigh market enough price to cover for the their grid. higher In Texas, costs. This highlights an important $/kWhgrid to generatelike LCOE .the additionalLACE is electricity the market that price would (MP be), whichrequired is the if that price power plant were not available t feature of LACE: for conventionalwind plants are power relatively plants, more LACE available does not during only depend on each particular (EIA 2016d). The word "avoided"that a generator is included would in receivethe LACE in acronymthe market to ifreflect it thethe cost night savings when associated market prices are low, while solar withOne interpretationavoiding the less of theefficientsells LACE one generationof unit a power of energy thatplant duringis is replaced that period. it istechnology, by a measure theIt is analyzed defined butof how also technology. plants,muchon the it formarket would example, costas a theproducewhole. For during example the day, a certain when technology in a market with high grid to generate the additionalas the highest electricity of the that dispatched would be prices requiredprices of wouldthe if that online have power a marketmuch plant higher were prices not LACE are available higher. than the Thus, same the tLACE echnology of a solarin a low -price market. Therefore, LACE (EIAThe first2016d) concept. The wordto understand "avoided"plants in before periodis included calculating t. A calculationin the LACELACE of isacronymis revenuethe not market an to mustappropriate reflect price the( plant waycost), which tends tosavings compare isto the beassociated higherprice technologies that than LACE across for markets. a wind reflect that fact that each power plant generates a plant; a solar plant allows the system to avoid more witha generator avoiding would the less receive efficient in the generation market if thatit sells is replacedone unit ofby energythe analyzed during technology. period𝑡𝑡 . It is defined as the different amount of electricity at different times. 𝑀𝑀𝑀𝑀 highest of the dispatched prices of the online plants in periodRenewable t. A calculation plants have ofexpensive revenue a marginal electricity must cost reflect of compared $0 /MWh, to and a wind so they plant. are always dispatched when available. The first concept to understand before calculating LACE isTherefore, the market their price LACE ( only), which𝑡𝑡 depends is the on price the marketthat price and the availability of the generator that that fact that each powerLACE plant can generates be calculated a different as the amount weighted of electricity at differentIn summary, times. LACE is the average value of a generator would receiveaverage in the of market the market if it sells price one of theunit powerdetermines of energy during the market period price. It foris defined the grid as. In the Texas, wind plants are relatively more available during the 𝑀𝑀𝑀𝑀revenue𝑡𝑡 per unit of energy sold in the market, highestLACE can of be the calculated dispatched assource pricesthe weighted that of theit generates online average plants (Namovicz of the in periodmarketnight 2013). t price.when A calculation of market the power pricesofignoring revenue source are the low, mustthat costs while it reflect of solargeneration. plants, Therefore, for example, produce during the day when market 𝑡𝑡 thatgenerates fact that (Namovicz each power 2013)(5) plant. generates a different amountprices of areelectricity higher. at Thus differentit cannot, the LACE times.be used of a independentlysolar plant tends to tomake be higher than LACE for a wind plant; a solar plant allows the systemcomparisons to avoid more across expensive different electricitymarkets. compared to a wind plant. LACE can be calculated as the weighted average of the market price of the power source(7 )that it 𝑇𝑇 In summary, LACE is Netthe averagevalue value of revenue per unit of energy sold in the market, ignoring the generates (Namovicz 2013). ∑𝑡𝑡 𝐴𝐴𝐴𝐴𝐺𝐺𝑛𝑛,𝑡𝑡 × 𝑀𝑀𝑃𝑃𝑡𝑡 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑛𝑛,𝑡𝑡 = 𝑇𝑇 costs of generation. Therefore, it cannot be used independently to make comparisons across different where is the actualwhere production AAG is of the technology actual∑ production𝑡𝑡 𝐴𝐴𝐴𝐴 n in𝐺𝐺𝑛𝑛 time,𝑡𝑡 of t. Noting that forAs a mentioned year, before, LACE and LCOE cannot n,t markets. be used independently(7) to determine the best technology n in time t. 𝑇𝑇Noting that for a year, 𝑇𝑇 𝑛𝑛,𝑡𝑡we can substitute into the denominator of (7) to obtain: 𝑡𝑡 𝑛𝑛,𝑡𝑡 𝐴𝐴𝐴𝐴𝐺𝐺 T × ∑𝑡𝑡 𝐴𝐴𝐴𝐴𝐺𝐺𝑛𝑛,𝑡𝑡 × 𝑀𝑀𝑃𝑃𝑡𝑡 generation∑ technology𝐴𝐴𝐴𝐴𝐺𝐺 = to construct or the Σt AAGn,t = 8,760𝑛𝑛,𝑡𝑡 CFn,t we𝑇𝑇 can substitute 𝑛𝑛,𝑡𝑡 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 = Net value worst technology to decommission. However, where8,760 × 𝐶𝐶𝐶𝐶 is the actualinto production the denominator of technology of (5)∑𝑡𝑡 to𝐴𝐴𝐴𝐴 n obtain: in𝐺𝐺𝑛𝑛 time,𝑡𝑡 t. Noting that for a year, As mentioned before,the LACE difference and(8) LCOE between cannotLACE be used and LCOEindependently, to determine the best generation (6) 𝑇𝑇 𝑇𝑇 we can substitute into the denominator𝑡𝑡 𝑛𝑛 ,of𝑡𝑡 (7) to𝑡𝑡 obtain: referred to as net value, accounts for revenues 𝐴𝐴𝐴𝐴𝐺𝐺𝑛𝑛,𝑡𝑡 ∑ 𝐴𝐴𝐴𝐴𝐺𝐺 ×technology𝑀𝑀𝑃𝑃 to construct or the∑ worst𝑡𝑡 𝐴𝐴𝐴𝐴 𝐺𝐺technology𝑛𝑛,𝑡𝑡 = to decommission. However, the difference between 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑛𝑛,𝑡𝑡 = and cost, making it possible to calculate the 𝑛𝑛,𝑡𝑡 𝑛𝑛,𝑡𝑡 8From,760 equation× 𝐶𝐶𝐶𝐶 (8) we can infer that a power plant8,760 which× 𝐶𝐶 𝐶𝐶isLACE always and available LCOE, referred butprofit is notto ofas dispatched each net valuetechnology ,would accounts (Namovicz for revenues 2013): and cost, making it possible to calculate (8) have a LACE of $0/MWh, because it does not sell𝑇𝑇 any unitsthe of profitelectricity, of each i.e. technology, (7) (Namovicz 2013): 𝑡𝑡 𝑛𝑛,𝑡𝑡 𝑡𝑡 𝑛𝑛,𝑡𝑡 ∑ 𝐴𝐴𝐴𝐴𝐺𝐺 × 𝑀𝑀𝑃𝑃 In order to be dispatched, a conventional 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 power= plant must𝑛𝑛 ,𝑡𝑡offer to generate𝐴𝐴𝐴𝐴 electricity𝐺𝐺𝑛𝑛,𝑡𝑡 = 0 at ∀ a𝑡𝑡 .price that is From equation (8) we canFrom infer equation that a power (6) we plant can8,760 infer which× that𝐶𝐶 𝐶𝐶is alwaysa power available plant but is not dispatched would (9) low enough to enter into the market; less efficient plants are dispatched only at periods of higher have a LACE of $0/MWhwhich, because is always it does available not sell but any is units not dispatchedof electricity, i.e., 𝑁𝑁𝑁𝑁𝑁𝑁𝑛𝑛,𝑡𝑡 × 𝐶𝐶𝐶𝐶𝐶𝐶𝑛𝑛,𝑡𝑡 𝑁𝑁𝑁𝑁𝑁𝑁 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑛𝑛,𝑡𝑡 = 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑛𝑛,𝑡𝑡 − 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑛𝑛,𝑡𝑡 = The net value can be also𝑛𝑛 ,𝑡𝑡calculated as the net present value (NPV)𝐴𝐴𝐴𝐴̅̅̅̅ ̅𝐺𝐺annualized̅̅𝑛𝑛̅,̅𝑡𝑡 with the CRF and divided In order to be dispatched, a conventional power plant must offer to generate𝐴𝐴𝐴𝐴 electricity𝐺𝐺 = 0 at ∀ a𝑡𝑡 .price that 73is by the annual average generation, . Therefore, if the net value is positive, the net present value low enough to enter into the market; less efficient plants are dispatched only at periods of higher of the project will be positive, and vice-versa. As a result, the net value can give us insight into which 𝐴𝐴𝐴𝐴̅̅̅̅̅𝐺𝐺̅̅𝑛𝑛̅,̅𝑡𝑡 The Full Cost of Electricity (FCe-) technologies are profitableModel at Documentation a given time and Results in 73a for given ERCOT marketScenarios, Julyand 2017 should | 54 therefore be expected to see increases in installed capacity relative to technologies with lower net values.

74

Thenet value can be also calculated as the net present value (NPV) annualized with the CRF and divided by the annual average generation, AAGn,t. Therefore, if thenet value is positive, the net present value of the project will be positive, and vice-versa. As a result, the net value can give us insight into which technologies are profitable at a given time in a given market and should therefore be expected to see increases in installed capacity relative to technologies with lower net values.

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 55 Appendix C: Wind Rating Factors by Aggregate County (%)

Annual June September County Load Zone Min Mean Max Min Mean Max Min Mean Max Andrews West 0.00 40.18 94.21 0.04 57.93 93.63 0.00 34.50 90.80 Archer West 0.00 42.52 95.87 0.00 60.63 95.00 0.00 28.04 95.20 Armstrong West 0.00 42.74 92.52 0.39 53.42 92.17 0.00 34.46 90.64 Borden West 0.00 38.98 94.17 0.00 54.49 92.80 0.00 27.27 90.49 Briscoe West 0.00 38.99 94.21 0.00 49.31 93.47 0.00 27.72 90.82 Callahan West 0.00 39.09 93.28 0.00 55.63 92.82 0.00 22.99 89.61 Cameron South 0.00 39.33 94.70 0.33 50.87 91.39 0.00 26.39 86.56 Carson West 0.00 44.92 94.62 0.00 57.42 94.02 0.00 40.19 92.69 Castro West 0.00 41.00 93.57 0.00 48.05 93.00 0.00 31.58 90.93 Childress West 0.00 39.20 96.49 0.00 55.20 93.81 0.00 29.78 92.53 Clay West 0.00 41.34 97.00 0.00 58.03 95.80 0.00 28.05 93.91 Coke West 0.00 35.48 95.44 0.00 49.55 94.40 0.00 22.45 84.48 Comanche North 0.00 42.59 95.74 0.00 59.07 93.97 0.00 22.39 93.35 Cooke North 0.00 37.51 96.98 0.00 50.56 95.73 0.00 22.85 95.91 Coryell North 0.00 40.64 94.52 0.02 55.93 92.87 0.00 21.09 89.98 Crockett West 0.00 40.94 93.17 0.00 64.20 92.86 0.00 31.86 91.45 Crosby West 0.00 41.49 96.55 0.00 54.69 95.34 0.00 29.37 90.90 Dallam West 0.00 44.63 93.50 0.21 55.23 93.19 0.00 40.88 91.49 Dawson West 0.00 38.85 93.50 0.00 54.51 92.17 0.00 27.10 88.94 Deaf Smith West 0.00 42.53 94.42 0.00 48.15 93.02 0.00 36.32 92.33 Dickens West 0.00 44.04 96.67 0.00 58.12 96.00 0.00 30.37 93.47 Eastland North 0.00 39.52 93.37 0.11 55.78 92.31 0.00 22.13 91.79 Erath North 0.00 30.29 97.00 0.00 44.59 94.17 0.00 14.75 90.00 Floyd West 0.00 41.34 93.68 0.00 53.50 92.65 0.00 29.52 92.06 Galveston (Offshore) Houston 0.00 30.76 97.00 0.00 23.53 80.54 0.00 20.74 94.60 Glasscock West 0.00 40.84 94.22 0.00 56.91 92.75 0.00 29.33 91.76 Gray West 0.00 44.94 93.32 0.00 55.98 93.24 0.00 38.23 91.79 Hale West 0.00 40.58 92.65 0.07 50.66 91.88 0.00 29.67 90.91 Hall West 0.00 39.03 93.98 0.00 50.53 93.24 0.00 28.14 90.24 Haskell West 0.00 40.53 94.65 0.00 57.61 93.63 0.00 28.13 92.17 Hidalgo South 0.00 36.98 95.26 0.00 49.22 91.92 0.00 23.89 89.08 Howard West 0.00 39.78 94.08 0.00 55.37 93.31 0.00 27.82 92.42 Jack North 0.00 39.63 94.45 0.00 54.36 92.80 0.00 24.04 91.72

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 56 Annual June September County Load Zone Min Mean Max Min Mean Max Min Mean Max Jim Hogg South 0.00 39.72 96.30 0.09 59.32 96.30 0.00 33.60 94.31 Kenedy South 0.00 33.21 93.67 0.11 46.45 93.05 0.00 25.18 87.86 Kent North 0.00 39.85 96.24 0.00 54.01 95.09 0.00 25.84 93.21 Kinney South 0.00 31.57 93.79 0.00 53.04 91.54 0.00 27.21 92.86 Kleberg South 0.00 39.04 95.10 0.25 51.11 93.00 0.00 31.15 89.46 Knox West 0.00 39.13 96.40 0.00 55.67 94.55 0.00 27.63 91.66 Live Oak South 0.00 33.62 96.80 0.00 49.07 95.90 0.00 27.61 92.99 Martin West 0.00 34.43 95.79 0.00 51.53 94.01 0.00 23.33 89.97 McCulloch South 0.00 41.91 94.21 0.04 57.86 92.75 0.00 23.72 91.54 Mills North 0.00 40.94 96.35 0.00 55.22 93.81 0.00 21.18 94.31 Mitchell West 0.00 37.92 95.27 0.00 55.14 93.87 0.00 27.92 93.57 Nolan West 0.00 39.99 92.91 0.00 56.31 92.71 0.00 26.08 90.49 Nueces South 0.00 38.44 93.59 0.24 50.67 92.73 0.00 30.77 91.19 Oldham West 0.00 44.26 95.19 0.00 52.33 93.88 0.00 41.79 92.15 Parmer West 0.00 40.78 94.08 0.00 46.39 92.25 0.00 31.22 91.30 Pecos West 0.00 43.44 93.29 2.65 69.20 93.29 0.00 43.16 92.11 Randall West 0.00 43.44 95.28 0.00 52.56 95.09 0.00 34.95 94.80 Reagan West 0.00 39.16 92.47 0.08 59.36 91.52 0.01 29.87 90.39 San Patricio South 0.00 38.27 93.90 0.24 50.54 93.05 0.00 30.66 91.69 Schleicher West 0.00 42.71 96.90 0.00 60.60 96.15 0.00 28.68 93.47 Scurry West 0.00 41.53 95.15 0.00 57.16 94.00 0.00 27.20 92.79 Shackelford West 0.00 39.00 94.20 0.00 55.36 93.60 0.00 22.78 91.58 Starr South 0.00 37.95 93.90 0.06 52.81 91.10 0.00 27.34 87.32 Sterling West 0.00 35.82 95.05 0.00 50.72 92.34 0.00 23.88 89.70 Stonewall West 0.00 37.71 96.99 0.00 51.53 96.46 0.00 24.85 96.16 Swisher West 0.00 40.41 93.89 0.00 49.99 91.98 0.00 29.73 91.31 Taylor West 0.00 39.14 93.87 0.00 55.80 93.40 0.00 23.12 91.41 Tom Green West 0.00 41.68 96.80 0.00 58.04 95.47 0.00 26.64 89.20 Upton West 0.00 36.59 93.63 0.00 61.50 93.24 0.00 31.89 91.10 Val Verde West 0.00 39.23 95.66 0.00 65.85 94.73 0.00 32.40 93.40 Webb South 0.00 37.62 93.74 0.00 59.19 92.98 0.00 30.27 91.84 Wharton Houston 0.00 37.69 93.85 0.26 50.35 92.63 0.00 30.28 91.53 Wilbarger West 0.00 43.09 94.70 0.00 60.92 93.53 0.00 31.44 93.72 Willacy South 0.00 33.40 93.65 0.03 44.72 90.90 0.00 22.30 82.63 Winkler West 0.00 44.84 96.92 0.00 63.12 96.92 0.00 43.55 96.53 Young North 0.00 36.81 96.13 0.00 52.98 95.16 0.00 22.93 90.44 Zapata South 0.00 38.13 93.22 0.04 59.22 93.20 0.00 31.08 91.34

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 57 Appendix D: ERCOT Hardwired Plant Additions for the CT Scenario

Variable Maximum Start Forced Net Average Fixed O&M O&M Minimum Ramp Minimum Minimum Cost Outage Prime Capacity Heat Rate Charge [$/ Charge Stable Rate [%/ Down Up Time [$/MW- Rate Generator County Load Zone Online Fuel Mover [MW] [Btu/kWh] kW-yr] [$/MWh] Level [%] min] Time [hrs] [hrs] start] [%]

Colorado Bend III Wharton Houston 2017 NG CCGT 1,148 7,468 18.00 3.19 32 0.53 10.7 5.7 23.80 0

DeCordova 5 & 6 Hood North 2018 NG CCGT 450 7,468 18.00 3.19 32 0.53 10.7 5.7 23.80 0

Wolf Hollow II Hood North 2017 NG CCGT 1,118 7,468 18.00 3.19 32 0.53 10.7 5.7 23.80 0

Antelope Station Hale West 2016 NG IC 168 8,500 22.72 4.73 30 25 4 4 52.00 0

Red Gate Power Plant Hidalgo South 2016 NG IC 225 8,500 22.72 4.73 30 25 4 4 52.00 0

Sky Global Power One Colorado South 2016 NG IC 51 8,500 22.72 4.73 30 25 4 4 52.00 0

Elk Station I Hale West 2016 NG OCGT 202 9,000 10.60 15.06 25 14 16 16 25.00 0

Elk Station II Hale West 2016 NG OCGT 202 9,000 10.60 15.06 25 14 16 16 25.00 0

Elk Station III Hale West 2016 NG OCGT 202 9,000 10.60 15.06 25 14 16 16 25.00 0

Lake Creek 3 McLennan North 2018 NG OCGT 450 9,000 10.60 15.06 25 14 16 16 25.00 0

P. H. Robinson Peaker Galveston Houston 2016 NG OCGT 388 9,000 10.60 15.06 25 14 16 16 25.00 0

Pecan Creek Energy Nolan West 2017 NG OCGT 270 9,000 10.60 15.06 25 14 16 16 25.00 0 Center

Castle Gap Solar Upton West 2016 Solar PV 116 - 25.79 0.00 - - - - - 0 Project

RE Roserock Solar Pecos West 2016 Solar PV 160 - 25.79 0.00 - - - - - 0

Riggins Solar Pecos West 2016 Solar PV 150 - 25.79 0.00 - - - - - 0 (SunEdison Buckthorn Westex, Oak Solar)

Solara Solar (OCI Alamo Haskell West 2016 Solar PV 106 - 25.79 0.00 - - - - - 0 7, Paint Creek)

West Texas Solar (OCI Pecos West 2016 Solar PV 110 - 25.79 0.00 - - - - - 0 Alamo 6)

Bearkat Renewable Glasscock West 2017 Wind WT 240 - 41.18 0.00 - - - - - 0 Energy Project

Changing Winds Castro West 2017 Wind WT 288 - 41.18 0.00 - - - - - 0

Colbeck’s Corner Wind Carson West 2016 Wind WT 200 - 41.18 0.00 - - - - - 0 Farm

Easter Renewable Castro West 2017 Wind WT 300 - 41.18 0.00 - - - - - 0 Energy Project

Fluvanna Renewable 1 Scurry West 2017 Wind WT 240 - 41.18 0.00 - - - - - 0

Goodnight Wind Energy Armstrong West 2017 Wind WT 240 - 41.18 0.00 - - - - - 0

Gunsight Mountain Howard West 2016 Wind WT 120 - 41.18 0.00 - - - - - 0 Wind

Horse Creek Wind Haskell West 2016 Wind WT 200 - 41.18 0.00 - - - - - 0

Los Vientos Wind IV Starr South 2016 Wind WT 200 - 41.18 0.00 - - - - - 0

Los Vientos Wind V Starr South 2016 Wind WT 110 - 41.18 0.00 - - - - - 0

Mariah del Norte Parmer West 2016 Wind WT 230 - 41.18 0.00 - - - - - 0

Mariah del Sur Parmer West 2017 Wind WT 230 - 41.18 0.00 - - - - - 0

Rock Springs Val Verde Val Verde South 2016 Wind WT 180 - 41.18 0.00 - - - - - 0 Wind

Salt Fork 1 Wind Gray West 2016 Wind WT 200 - 41.18 0.00 - - - - - 0

Sendero Wind Energy Jim Hogg South 2016 Wind WT 78 - 41.18 0.00 - - - - - 0

South Plains Wind Floyd West 2016 Wind WT 152 - 41.18 0.00 - - - - - 0 Phase II

Texas Wind Farm Haskell West 2017 Wind WT 400 - 41.18 0.00 - - - - - 0

Wake Wind Dickens West 2016 Wind WT 300 - 41.18 0.00 - - - - - 0

Baffin Wind (Penascal Kenedy South 2016 Wind-C WT 202 - 41.18 0.00 - - - - - 0 3)

San Roman Wind Cameron South 2016 Wind-C WT 103 - 41.18 0.00 - - - - - 0

South Texas Wind Farm Kenedy South 2016 Wind-C WT 200 - 41.18 0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 58 Appendix E: ERCOT Hardwired Plant Additions for the AR Scenario (in addition to the CT scenario)

Average Fixed Variable Heat O&M O&M Maximum Minimum Start Forced Net Rate Charge Charge Minimum Ramp Down Minimum Cost Outage Prime Capacity [Btu/ [$/kW- [$/ Stable Rate [%/ Time Up Time [$/MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] Level [%] min] [hrs] [hrs] start] [%]

Andrews 7 Solar Andrews West 2016 Solar PV 80 - 25.79 $0.00 - - - - - 0

Austin Community Solar Travis South 2016 Solar PV 3.2 - 25.79 $0.00 - - - - - 0

Barilla Solar 1B Pecos West 2016 Solar PV 7 - 25.79 $0.00 - - - - - 0

Barilla Solar 2 Pecos West 2016 Solar PV 21 - 25.79 $0.00 - - - - - 0

Bluebell Solar Sterling West 2016 Solar PV 173 - 25.79 $0.00 - - - - - 0

BNB Lamesa Solar Dawson West 2016 Solar PV 200 - 25.79 $0.00 - - - - - 0

Brewster Solar Project Brewster West 2016 Solar PV 30 - 25.79 $0.00 - - - - - 0

Capricorn Ridge Solar Coke West 2016 Solar PV 100 - 25.79 $0.00 - - - - - 0

Center for Solar Energy Bell North 2018 Solar PV 50 - 25.79 $0.00 - - - - - 0

East Pecos Solar Pecos West 2016 Solar PV 100 - 25.79 $0.00 - - - - - 0

Horseshoe Bend Solar Project Parker North 2016 Solar PV 140 - 25.79 $0.00 - - - - - 0

Kingsberry Community Solar Travis South 2016 Solar PV 3.2 - 25.79 $0.00 - - - - - 0

Mesquite Solar Project Tom Green West 2016 Solar PV 10 - 25.79 $0.00 - - - - - 0

Nazareth Solar Castro West 2016 Solar PV 201 - 25.79 $0.00 - - - - - 0

Pearl Solar (OCI Alamo 6 Pecos West 2016 Solar PV 50 - 25.79 $0.00 - - - - - 0 Phase B)

Pecos County Solar Project Pecos West 2018 Solar PV 170 - 25.79 $0.00 - - - - - 0 (Hanwha)

Pecos Solar Power I Pecos West 2017 Solar PV 108 - 25.79 $0.00 - - - - - 0

Pflugerville Solar Farm (RRE Travis South 2017 Solar PV 60 - 25.79 $0.00 - - - - - 0 Austin Solar)

SolaireHolman 1 Brewster West 2016 Solar PV 50 - 25.79 $0.00 - - - - - 0

Synergy Community Solar Grimes North 2016 Solar PV 1.2 - 25.79 $0.00 - - - - - 0 Project

Unity Solar Project Deaf Smith West 2018 Solar PV 100 - 25.79 $0.00 - - - - - 0

Upco Power 1 (SP-TX-12) Upton West 2016 Solar PV 180 - 25.79 $0.00 - - - - - 0

Upco Power 2 (SP-TX-12 Upton West 2016 Solar PV 120 - 25.79 $0.00 - - - - - 0 Phase B)

Upton Solar Upton West 2017 Solar PV 102 - 25.79 $0.00 - - - - - 0

White Camp Solar Farm Kent West 2016 Solar PV 102.02 - 25.79 $0.00 - - - - - 0

Albercas Wind Zapata South 2016 Wind WT 250 - 41.18 $0.00 - - - - - 0

Buckthorn Wind 1 Erath North 2016 Wind WT 96 - 41.18 $0.00 - - - - - 0

Cameron Wind Phase 2 Cameron South 2019 Wind-C WT 150 - 41.18 $0.00 - - - - - 0

Canyon Wind Project Scurry West 2017 Wind WT 300 - 41.18 $0.00 - - - - - 0

Caprock Wind Castro West 2017 Wind WT 300 - 41.18 $0.00 - - - - - 0

Chapman Ranch Wind I Nueces South 2016 Wind-C WT 250 - 41.18 $0.00 - - - - - 0

Comanche Ridge Project Stonewall West 2018 Wind WT 24 - 41.18 $0.00 - - - - - 0

Comanche Run Wind Swisher West 2016 Wind WT 500 - 41.18 $0.00 - - - - - 0

Cotton Plains Wind 1 (Blanco Floyd West 2016 Wind WT 50 - 41.18 $0.00 - - - - - 0 Canyon)

Cotton Plains Wind 2 (Blanco Floyd West 2016 Wind WT 150 - 41.18 $0.00 - - - - - 0 Canyon)

Crosby County Wind Farm Crosby West 2016 Wind WT 150 - 41.18 $0.00 - - - - - 0

Dallam Ranch Wind Dallam West 2018 Wind WT 300 - 41.18 $0.00 - - - - - 0

Electra Wind Wilbarger West 2016 Wind WT 230 - 41.18 $0.00 - - - - - 0

Espiritu Wind Cameron South 2019 Wind-C WT 150 - 41.18 $0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 59 Average Fixed Variable Heat O&M O&M Maximum Minimum Start Forced Net Rate Charge Charge Minimum Ramp Down Minimum Cost Outage Prime Capacity [Btu/ [$/kW- [$/ Stable Rate [%/ Time Up Time [$/MW- Rate Generator County Load Zone Online Fuel Mover [MW] kWh] yr] MWh] Level [%] min] [hrs] [hrs] start] [%]

Galveston-Offshore Wind Galveston Houston 2018 Wind-O WT 300 - 41.18 $0.00 - - - - - 0

Grandview Wind Farm Carson West 2016 Wind WT 188 - 41.18 $0.00 - - - - - 0 Phase 3

Gulf Wind III Kenedy South 2016 Wind-C WT 187.2 - 41.18 $0.00 - - - - - 0

Hale Community Energy I Hale West 2016 Wind WT 136.35 - 41.18 $0.00 - - - - - 0

Hale Community Energy II Hale West 2019 Wind WT 363.6 - 41.18 $0.00 - - - - - 0

Happy Whiteface Wind Deaf Smith West 2016 Wind WT 157 - 41.18 $0.00 - - - - - 0

Haynes Wind Farm Gray West 2020 Wind WT 200 - 41.18 $0.00 - - - - - 0

Hidalgo & Starr Wind Hidalgo South 2016 Wind WT 250 - 41.18 $0.00 - - - - - 0

Live Oak Wind Project Schleicher West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Longhorn Wind South Briscoe West 2016 Wind WT 160 - 41.18 $0.00 - - - - - 0

Magic Valley Wind (Redfish) Willacy South 2017 Wind-C WT 115 - 41.18 $0.00 - - - - - 0 2A

Magic Valley Wind (Redfish) Willacy South 2017 Wind-C WT 115 - 41.18 $0.00 - - - - - 0 2B

Mariah del Este Parmer West 2017 Wind WT 139 - 41.18 $0.00 - - - - - 0

M-Bar Wind Andrews West 2020 Wind WT 200 - 41.18 $0.00 - - - - - 0

Miami Wind G3 Gray West 2016 Wind WT 111 - 41.18 $0.00 - - - - - 0

Midway Farms Wind San Patricio South 2016 Wind-C WT 161 - 41.18 $0.00 - - - - - 0

Palo Alto Farms West Wind Nueces South 2021 Wind-C WT 300 - 41.18 $0.00 - - - - - 0 Project

Palo Duro Wind Deaf Smith West 2016 Wind WT 203 - 41.18 $0.00 - - - - - 0

Pampa Wind Gray West 2017 Wind WT 500 - 41.18 $0.00 - - - - - 0

Panhandle Wind Ph 3 Carson West 2016 Wind WT 248 - 41.18 $0.00 - - - - - 0

Patriot Wind (Petronilla) Nueces South 2016 Wind-C WT 180 - 41.18 $0.00 - - - - - 0

Pullman Road Wind Randall West 2016 Wind WT 300 - 41.18 $0.00 - - - - - 0

Rattlesnake Den Wind 2 Glasscock West 2017 Wind WT 158 - 41.18 $0.00 - - - - - 0

RTS Wind McCulloch South 2016 Wind WT 200 - 41.18 $0.00 - - - - - 0

Salt Fork 2 Wind Carson West 2016 Wind WT 200 - 41.18 $0.00 - - - - - 0

Santa Rita Wind Reagan West 2016 Wind WT 300 - 41.18 $0.00 - - - - - 0

Scandia Wind Ph D (Mariah) Parmer West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Scandia Wind Ph E (Mariah) Parmer West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Scandia Wind Ph F (Mariah) Parmer West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Silver Canyon Wind Farm Briscoe West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Smart Wind Ranch (Spinning Upton West 2022 Wind WT 615 - 41.18 $0.00 - - - - - 0 Star)

South Plains Wind Phase III Floyd West 2016 Wind WT 148 - 41.18 $0.00 - - - - - 0

Stella Wind Farm Kenedy South 2016 Wind-C WT 200 - 41.18 $0.00 - - - - - 0

Stella Wind Farm II Kenedy South 2017 Wind-C WT 200 - 41.18 $0.00 - - - - - 0

Swisher Wind Swisher West 2016 Wind WT 300 - 41.18 $0.00 - - - - - 0

Tecovas Wind Project Briscoe West 2017 Wind WT 200 - 41.18 $0.00 - - - - - 0

Torrecillas Wind A Webb South 2016 Wind WT 200 - 41.18 $0.00 - - - - - 0

Torrecillas Wind B Webb South 2016 Wind WT 200 - 41.18 $0.00 - - - - - 0

Tyler Bluff Wind (Muenster) Cooke North 2016 Wind WT 118 - 41.18 $0.00 - - - - - 0

Unity Wind Deaf Smith West 2016 Wind WT 203 - 41.18 $0.00 - - - - - 0

Wharton Wind Project Wharton Houston 2018 Wind WT 250 - 41.18 $0.00 - - - - - 0

Willow Springs Wind Haskell West 2016 Wind WT 200 - 41.18 $0.00 - - - - - 0

The Full Cost of Electricity (FCe-) Model Documentation and Results for ERCOT Scenarios, July 2017 | 60