Positioning of Danish offshore wind farms until 2030 – using Levelized Cost of Energy (LCoE)

Wind Energy Master Report Department of

Gyde Liane Ohlsen

DTU Wind Energy-M-0262

January 2019

Author: Gyde Liane Ohlsen DTU Wind Energy-M-0262 Title: Positioning of Danish offshore wind farms until 2030 – using January 2019 Levelized Cost of Energy (LCoE)

ISBN: 978-87-93549-50-0 Project period: August 2018 – January 2019

ECTS: 30

Education: Master of Science

Supervisors:

Niels-Erik Clausen Asger Bech Abrahamsen DTU Wind Energy

Remarks: This report is submitted as partial fulfillment of the requirements for graduation in the above education at the Technical University of .

DTU Wind Energy is a department of the Technical University of Denmark with a unique integration of research, education, innovation and public/private sector consulting in the field of wind energy. Our activities develop new Technical University of Denmark opportunities and technology for the global and Danish exploitation of wind Department of Wind Energy energy. Research focuses on key technical-scientific fields, which are central Frederiksborgvej 399 2800 Kgs. Lyngby for the development, innovation and use of wind energy and provides the Denmark basis for advanced education at the education. www.vindenergi.dtu.dk

We have more than 240 staff members of which approximately 60 are PhD students. Research is conducted within nine research programmes organized into three main topics: Wind energy systems, technology and Basics for wind energy. c d Abstract

The cost of offshore wind energy projects has declined significantly during the past decade. A key element for this reduction in project costs has been the identification of the most attractive areas that maximize the business case value of an offshore project, from the point of view of both government and project developer. This Master Thesis presents a methodology that allows a comparison between potential offshore wind farm project locations using Levelized Cost of Energy (LCoE) - an indicator representing the cost of one unit of produced energy along the lifetime of the project. Predictions of future cost and technology development up until 2030 are introduced in order to identify factors that have potential to reduce LCoE in the future. The most profitable areas were found around the coast of Jutland in the NorthSea and Kattegat (northwest of Jutland). However, also sites east of Sealand appeared to be attractive. An overall drop in LCoE across the entire Danish Sea from 115-200 e/MWh in 2008 to 36-60 e/MWh in 2030 was determined. This drop was caused by capital and operational expenditure (CAPEX and OPEX) reductions, but as well by technology improvements such as increased turbine and farm capacity as well as extended lifetime. Furthermore, a sensitivity analysis revealed that the discount rate, the wind turbine costs and the OPEX have significant impact on the LCoE. A ranking according to the LCoE of wind farms positioned in pre-selected areas [1] was performed for 2021 and compared to an external report prepared by COWI A/S [2], showing LCoE values in the same order of magnitude. With a LCoE of 53 e/MWh the contained the most profitable areas (Jammerbugt and Nordsøen), followed by the Kriegers Flak area with 54 e/MWh and Hesselø with 58 e/MWh. The close range of these values revealed, that the reduction in LCoE might not necessarily depend on the site selection but rather has to result from a reduction in cost or improvement in technology. Finally, the obtained range of LCoE for future offshore wind until 2030 was com- pared to electricity price predictions [3] to discuss the possibility of a subsidy-free future for offshore wind energy projects. It was concluded that this might be pos- sible if electricity prices will increase notable during the coming years. However, if prices follow the low electricity price scenario, a subsidy-free future is believed hard to obtain - not even when considering a low LCoE scenario. ii Preface

This Master thesis was prepared at the department of Wind Energy of DTU in Risø in fulfillment of the requirements for acquiring a Master of Science degree in Sustainable Energy with a specialization in Wind Energy.

DTU Risø Campus, Roskilde, March 12, 2019

Gyde Liane Ohlsen, (s161887) iv Acknowledgement

First of all, I would like to thank my supervisors, Niels-Erik Clausen and Asger Bech Abrahamsen, for the constant support during the past five month. I am very grateful for all your feedback, interesting discussions and the time you have spend along with the work of this Thesis.

Secondly, I would like to acknowledge the WAsP and GWA team, including Niels Gylling Mortensen, Duncan Heathfield and Morten Nielsen, for providing me with all kinds of information and data related to wind resources and WAsP and even answer- ing my emails on a Sunday night.

Also I would like thank the Technical University of Denmark and especially the department of DTU Wind Energy in Risø, for giving me the opportunity to develop my academic career in their educational system.

Special mention goes to Guillermo, for always being there for me, calming me down when I get stressed and always supporting me. Thank you for everything.

Last but not least, I thank my friends and especially my family. Without your support I would never have been where I am right now. Thank you Antje, Hennes and Flemming! vi Table of Contents

Abstract i

Preface iii

Acknowledgement v

Table of Contents vii

List of Figures viii

List of Tables xii

1 Introduction 1 1.1 Motivation ...... 2 1.2 Literature review ...... 4 1.3 Objective ...... 5 1.4 Content ...... 6

2 Wind turbine theory 7 2.1 Utilized power curves ...... 9

3 Methodology 11 3.1 Levelized Cost of Energy (LCoE) ...... 11 3.2 Assumptions and simplifications ...... 12 3.3 Model setup ...... 13

4 Input data 15 4.1 Wind resources and Annual Energy Production (AEP) ...... 15 4.2 Water depth ...... 18 4.3 Capital expenditures (CAPEX) ...... 19 4.4 Operational expenditures (OPEX) ...... 31 4.5 Other expenses ...... 32 4.6 Wind farm losses ...... 32

5 Results 37 5.1 Cost of Energy (CoE) ...... 38 5.2 Calibration of discount rate ...... 39 5.3 Levelized Cost of Energy (LCoE) ...... 42 5.4 Uncertainty and sensitivity ...... 48 5.5 Selected sites ...... 52 5.6 Electricity price comparison ...... 62

6 Conclusion 65

7 Outlook 67

A Additional graphs 69 A.1 Site selection ...... 69 A.2 Transmission ...... 70

References 71

List of Figures

1.1 Predicted cumulative operating and annually installed capacity for Eu- ropean offshore wind energy from 2016-2045 (reproduced from [5], with permission from the publisher IRENA)...... 1 1.2 Map of the Danish Sea showing the pre-selected sites by the DEA [1] (pur- ple line) and 5 existing/planned offshore wind farms (red dots) (Anholt [10], 3 (HR 3) [11], Kriegers Flak (KF) [12], Vesterhav (VH) Nord [13] and Syd [14]). Boundary of the Danish Sea (black dashed line). 3

2.1 Basic setup of an offshore wind farm including the offshore substation. The turbines are installed on substructures consisting of monopile and transition piece...... 7 2.2 Power curve (left y-axis scale) and thrust coefficient curve (right y-axis scale) of the Siemens wind turbine SWT-3.6-120 [29] with rated power of 3.6 MW and a rotor diameter of 120 m. Air density ρ equals 1.225 kg·m−3. (WAsP input file)...... 9 List of Figures ix

2.3 Power curves for the Siemens wind turbine (SWT-3.6MW-120) [29] with rated power of 3.6 MW and a rotor diameter of 120 m, the (V164- 8.0MW) [33] with rated power of 8 MW and rotor diameter 164 m, the DTU reference turbine (DTU 10MW-RWT) [31] with 10 MW rated power and 178 m rotor diameter and a turbine with rated power of 12 MW and a rotor diameter of 220 m (up-scaled from DTU 10 WM turbine)...... 10

3.1 Model setup, including LCoE-map boundaries and border of the Danish Sea (black dashed line) (left) and exemplary wind farm modelling around one pixel (right). UTM Zone 32N projection...... 13

4.1 Mean wind speed map of the Danish Sea, including boundary of the Danish Sea (black dashed line) and selected sites by the DEA [1] (purple line). .. 16 4.2 Weibull probability density function (red) with k = 2 and A = 10 ms−1. Power curve of the Vestas wind turbine V164-8.0MW [33] (blue)...... 17 4.3 Water depth of the Danish Sea, excluded depth >50 m (marked dark grey) and distance to shore <10 km (marked light grey)...... 18 4.4 Wind turbine price trends from 1997-2017 reproduced from [38] with per- mission from publisher IRENA, added a linear regression curve manually from 2008-2017 (black line), used in this study...... 20 4.5 Wind turbine cost development from 2008-2030. 2008-2017 obtained through [38], 2017-2030 learning rate (LR) of 10% and therefore a cost reduction of 16% is applied...... 21 4.6 Substructure cost as function of water depth, developed by DTU - Depart- ment of Wind Energy in 2015 [27], reproduced for this study...... 22 4.7 Substructure cost development at 20 m water depth from 2008-2030. 2008- 2017 linear regression obtained through past data from Kooijman [41], Douglas-Westwood [42], Danish Energy Agency [43] and NREL [44]. 2017- 2030 learning rate (LR) of 10%, thus cost reduction of 16% applied. ... 23 4.8 Exemplary wind farm inter-array cable layout including substation and transmission cable...... 25 4.9 Upper: high voltage alternating current (HVAC) transmission system setup, Lower: high voltage direct current (HVDC) transmission system setup [51]...... 27 4.10 HVDC cable cost over steady state current with voltage of 300 kV. Added polynomial regression curve to prices given by [51]...... 29 4.11 HVAC vs. HVDC electrical equipment cost for an exemplary wind farm with a capacity of 800 MW, including cost for onshore and offshore sub- station and transmission cable. Break-even length at approximately 97 km...... 30 4.12 OPEX development. Cost scatter up to 2015 derived from various sources: NREL [53], Ernst & Young [54], Engels [55], INNWIND [32], IEA [56]. OPEX prediction obtained by the Danish Energy Agency and [6]. Predictions for comparison given by BVG Associate [15]...... 32 x List of Figures

4.13 Wind rose and Weibull distribution at 100 m height (WAsP derived). Data obtained from the Global Wind Atlas [34] at position UTM Zone 32 (422042, 6206784) placed 20 km west of the Danish coast of Jutland. With Weibull A and k weighted for all sectors, U mean wind speed and P mean power density...... 33 4.14 Wind farm layout and wake losses for a 800 MW farm consisting of 100 -8.0MW turbines placed 20 km west of the Danish coast of Jutland (Fuga derived)...... 34 4.15 Wake losses depending on turbine spacing (6, 7 and 8×rotor diameter (dR)) and turbine type (Siemens SWT-3.6-120, Vestas V164-8.0MW and DTU 10MW RWT)...... 34

5.1 Cost of Energy (CoE)-map for the first 800 MW wind farm, tendered in 2019/2020 and assumed year of FID (Final investment decision) in 2021. Input parameters: Vestas V164-8.0MW turbine, lifetime of 30 years, AEP (annual energy production) determined using method in section 4.1, CAPEX (capital expenditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12. Areas closer than 10 km to shore marked light grey, areas with water depth deeper than 50 m marked dark grey, boundary of the Danish Sea indicated through black dashed line and the selected sites by the Danish Energy Agency (DEA) [1] surrounded by purple line...... 39 5.2 Development of the discount rate from 2008-2030 with uncertainty, includ- ing model derived discount rates (table 5.1) of existing or planned wind farms (Anholt [10], Horns Rev 3 [11], Kriegers Flak [12], Vesterhav Syd [14] and Nord [13]), COWI’s assumed rate of 8% [2], DEA’s of 4.5% [6] and IRENA’s discount rate [5]...... 42 5.3 Levelized Cost of Energy (LCoE)-map for the first 800 MW wind farm, tendered in 2019/2020 and assumed year of FID (final investment deci- sion) in 2021. Input parameters: Vestas V164-8.0MW turbine, lifetime of 30 years, AEP (annual energy production) determined using method in section 4.1, CAPEX (capital expenditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12 and the discount rate from figure 5.2. Areas closer than 10 km to shore marked light grey, areas with water depth deeper than 50 m marked dark grey, boundary of the Danish Sea indicated through black dashed line and the selected sites by the Danish Energy Agency (DEA) [1] surrounded by a purple line...... 43 5.4 Relative frequency and cumulative relative frequency of occurrence of LCoE across the Danish Sea (total considered area 51,293.25 km2) in the year 2021 of final investment decision (FID). (b) showing the 4 quartiles Q1, Q2 (median), Q3 and Q4 (maximum), meaning respectively 25%, 50%, 75% and 100% ≤ the respective LCoE...... 45 List of Figures xi

5.5 LCoE range across the Danish Sea from 2008-2030, considering different wind farm design developments, with the variables wind farm capacity PF , wind turbine capacity PT and lifetime LT . The bars for each year show the 4 quartiles Q1, Q2 (median), Q3 and Q4 (maximum), meaning respectively 25%, 50%, 75% and 100% ≤ the respective LCoE. The upper and lower boundaries of LCoE indicated by black line...... 47 5.6 LCoE uncertainty for baseline, high cost and low cost scenario in 2030, considering CAPEX ± 10%, OPEX ± 10% and discount rate ± 33%. Relative occurrence frequency of LCoE across the Danish Sea (left) and quartile analysis (right)...... 49 5.7 Contribution of each wind farm element to LCoE in the year of final in- vestment decision (FID) in 2021 (wind farm used for sensitivity analysis). Input parameters can be found in table 5.3, cost parameters in chapter 4: CAPEX (capital expenditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12, AEP (annual energy production) determined using method in section 4.1...... 50 5.8 LCoE sensitivity to cost parameters, discount rate and lifetime with the specifications stated in table 5.3 and final investment decision in2021. Cost parameters can be found in chapter 4: CAPEX (capital expendi- tures) derived from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12 and the discount rate from figure 5.2...... 51 5.9 Nordsøen wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived)...... 54 5.10 Nordsøen wind farm layout (dots) and disregarded zone (black dashed area) due to planned oil and gas pipelines and electrical interconnection between Denmark and the UK (Viking Link). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km...... 54 5.11 Jammerbugt wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived)...... 56 5.12 Jammerbugt wind farm layout (dots), original area (orange line) published by the DEA [1] and updated area (grey polygon) [2]. Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km...... 56 5.13 Hesselø wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived)...... 57 5.14 Hesselø wind farm layout (dots). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km...... 58 5.15 Kriegers Flak A wind rose (left) and Weibull probability density function (right) at 100 m height. (For Kriegers Flak B see Appendix A.1) ..... 59 5.16 Kriegers Flak wind farm layout (dots) for zone A (53 turbines) and zone B (47 turbines). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km. 59 5.17 Electricity price prediction including uncertainty reproduced from [3] com- pared to the LCoE for offshore wind across the Danish Sea from 2018-2030 presented in figure 5.5. The years of tender (triangle), expected finalin- vestment decision (FID) (circle) and expected year of commissioning (5 years after tender) (square) for the three planned offshore wind farms (1st farm ’red’, 2nd farm ’orange’, 3rd farm ’green’) shown in lower part of the figure...... 63

A.1 Kriegers Flak B wind rose (left) and Weibull probability density function (right) at 100 m height...... 69 A.2 Danish electricity grid, with marked strong points of common coupling of 400 kV (yellow circles)...... 70

List of Tables

3.1 Input parameters: Technical and economical aspects...... 14

4.1 Turbine and substructure installation time and installation vessel cost. .. 24

4.2 Inter-array cable key parameters with Vinter−array inter-array cable volt- age, ACu cross-section of cable, Issn [A] steady state current and Cinter−array cost per length of the inter-array cable...... 25 4.3 Suitable 400 kV onshore points of common coupling (PCC)...... 26

4.4 HVAC sub sea cable key parameters with VHV AC HVAC nominal voltage, Issn,HV AC as steady state current, ACu cross-section of cable, C capaci- tance per length of the cable and Ccable,HV AC HVAC cable cost per length. 28 4.5 prediction of [6] and the resulting OPEX determined by assuming 75% fixed and 25% variable cost and 8760 hours per year. ... 31 4.6 Wind farm AEP losses...... 35

5.1 Key parameters of the Danish wind farms Anholt, Horns Rev 3 (HR3), Kriegers Flak (KF), Vesterhav (VH) Nord and Syd (until first dashed line), including obtained wind resource data, water depth, distance to PCC (point of common coupling), CAPEX, OPEX and Net AEP (second part, from first dashed line to second). All calculations are performed with a lifetime of 25 years, a farm layout using turbine spacing of 7×rotor diameter and wind resource data at 100 m height and 80 m (Anholt only). LCoE given by [60] utilized to determine the discount rate (last part). .. 41 List of Tables xiii

5.2 Technological development from 2008-2030. Changed parameter indicated in bold font...... 46 5.3 Key parameters of the wind farm considered for the sensitivity analysis in final investment decision (FID) year 2021. AEP (annual energy pro- duction) determined using method in section 4.1. PCC (point of common coupling to grid)...... 50 5.4 Nordsøen offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Idomlund). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area...... 55 5.5 Jammerbugt offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Ferslev). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area...... 57 5.6 Hesselø offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Trige). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area. .. 58 5.7 Kriegers Flak A and B offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Bjæverskov). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area...... 60 5.8 Ranking of selected sites according to LCoE in final investment decision year 2021 and compared with COWI’s ranking in [2]...... 60 xiv CHAPTER 1 Introduction

Offshore wind has developed greatly during the last decade in Europe, becomingan economically attractive renewable energy option for governments and institutions to decrease greenhouse gas emissions. The main reasons for this remarkable growth have been first, major breakthroughs and developments on the technology, supply chain and overall industry and secondly due to the scarcity of good locations for land-based wind turbines. Offshore wind energy is expected to expand significantly in the future asfigure 1.1 illustrates. Offshore sites offer the possibility to install a great number of turbines under stable wind conditions, allowing the production of a large amount of green energy. Generally, the offshore technology trends move towards larger wind farms with greater turbine capacities, providing the potential to deliver an extensive and cost efficient contribution to the green transition [4].

Figure 1.1: Predicted cumulative operating and annually installed capacity for Eu- ropean offshore wind energy from 2016-2045 (reproduced from[5], with permission from the publisher IRENA).

However, to achieve a successful transition towards a fossil fuel free energy system, renewable energy needs to be competitive compared to energy generated by fossil fu- 2 1 Introduction els. To be able to compare the cost of energy provided by different generating units on a consistent basis, the Levelized cost of Energy (LCoE) is a widely used indicator. It represents the cost of one unit of produced energy (for instance given as e/MWh), by taking into account the investment and operating cost as well as the energy pro- duction over the entire lifetime of an energy project.

The LCoE for offshore wind energy has declined significantly during the past years in European countries mainly due to constant development in the technology, supply chain, construction and operation and maintenance (O&M) areas. The growing off- shore wind energy market has opened up for more competition, but at the same time the gained experience has led to an improved collaboration between the actors in the industry [6]. Offshore wind has also benefited from supporting governmental policies, decreasing the uncertainty and risk and thus lowering the cost of capital. Overall, offshore wind energy has become an attractive technology for governments aswell the private sector in order to fulfill the national and international climate goals.

The cost reduction for offshore wind projects was especially experienced in Den- mark. Here the LCoE has declined by nearly 50% from 2010 until 2016 [7]. The country started out in 1991 by erecting the world’s first offshore wind farm in the Danish inner waters, Vindeby. Since that time Denmark has successfully installed 13 offshore wind farms comprising an installed capacity of nearly 1.3 GW. Another four offshore wind farms are to be connected to the grid by 2020 with an additional capac- ity of 1.35 GW. Danish offshore wind projects are usually awarded through a public tendering process in order to construct new offshore wind farms at the lowest cost possible. The government invites project developers to submit their bids for a price at which they are willing to produce electricity. In general, the more challenging the conditions at offshore sites are (for instance low wind conditions with deep waters), the higher the bid prices will be. Thus, it is important for the tenderer that the area selected to be developed is the best possible to minimize project development costs (and hopefully also bids) and thereby maximizing society’s utility.

1.1 Motivation

In June 2018 the new Danish energy agreement was signed by the Danish government and all political parties. The aim of the agreement is to cover 55% of the Danish energy consumption and 100% of the electricity consumption by renewable energy sources in 2030 [8]. In 2017 renewable energy covered approximately 36% of the Danish energy consumption and 64% of electricity consumption [9]. Among others, it has been decided to expand the offshore wind energy generation capacity by erecting three new offshore wind farms until 2030 with a capacity ofat least 800 MW each. The goal is to include the grid connection costs in future tenders, which are nowadays covered by the transmission system operator (TSO), Energinet. The first of these wind farms will be tendered in 2019-2020 and commissioned between 1.1 Motivation 3

2024 and 2027. The second wind farm will be tendered in 2021 and the third in 2023, both planned to be operational before 2030. It is assumed that future offshore wind energy can participate in electricity markets without any external subsidies [8]. The agreement rose the question of where to erect those offshore wind farms. In order to identify potential areas for future offshore wind farms, the Danish Energy Agency (DEA) performed a large-scale screening of the Danish Sea. The result of this screening process was published in September 2018 by the DEA [1]. The pre-selected areas are shown in figure 1.2 together with the most recently installed and planned Danish offshore wind farms.

Figure 1.2: Map of the Danish Sea showing the pre-selected sites by the DEA [1] (purple line) and 5 existing/planned offshore wind farms (red dots) (Anholt [10], Horns Rev 3 (HR 3) [11], Kriegers Flak (KF) [12], Vesterhav (VH) Nord [13] and Syd [14]). Boundary of the Danish Sea (black dashed line).

The screening resulted in four areas - one in the North Sea (Nordsøen), one in the Jammerbugt, one close to Hesselø and one around Kriegers Flak. After detecting these pre-selected areas, it is important for the Danish government to identify the areas that are most attractive economic-wise in order to receive the most competitive bids during the planned tendering process. The Levelized Cost of Energy (LCoE) is a powerful indicator to use when comparing different sites according to profitability, thereby taking factors such as water depth, distance to shore and wind resources into account. 4 1 Introduction

The scope of this study is to provide a methodology to visualize and analyze the LCoE for offshore wind across the Danish Sea in order to identify the most profitable areas.

1.2 Literature review

In this section the latest and most relevant literature regarding LCoE studies for offshore wind energy will be reviewed. Several studies have undertaken analysis on LCoE for offshore wind energy in the past decade, some of them considering different price and technology development scenarios. However only a few provide a LCoE visualization to be able to compare the attractiveness of different offshore sites. Con- sidering the significant drop in LCoE for offshore wind during the last years, the publication year of these studies is decisive for the relevance for this study.

Different LCoE studies for European waters have been conducted. The most rele- vant and recent one regarding prediction of renewable energy generation in Denmark was published by the Danish Energy Agency (DEA) and Energinet in 2016 [6]. The section regarding offshore wind was updated in 2017 due to the notable changesin tender bidding prices. The study investigates the LCoE development up until 2050 by predicting future development of capital expenditure (CAPEX) and operational expenditure (OPEX) through available data from the industry. BVG Associate provides in 2017 a cost model investigating the impact of technol- ogy innovations on LCoE reductions for offshore wind energy from 2017 to 2030 in [15]. According to this study the main cost reduction is associated to an increase in turbine power rating up to 12 MW in 2030. But also improvements on the substruc- ture design and manufacturing, electrical design, the construction and development as well as in the operation, maintenance and service are expected to cause significant LCoE reductions. The DEA in [6] does not take varying site conditions, such as water depth, wind resources nor distance to shore into consideration, whereas [15] defines two sites with different conditions. However, both studies do not visualize the LCoE distribution across the sea.

The following studies provide LCoE variations for offshore wind shown graphically by usage of color maps. Though only Möller et al. in 2012 [16] presents a LCoE-map particularly for the Danish Sea. This study evaluates the meaning of scale by carrying out resource-economic analyses for two scenarios; one where the installation of offshore wind energy follows the trend, another one where the same amount of wind energy is produced in smaller, locally owned wind farms near shore. In 2017 BVG Associates and Geospatial Enterprises (GeoSE) prepared a study for WindEurope [17]. The report presents LCoE calculations for the Baltic Sea, the North Sea and the Atlantic taking into account two policy scenarios by the end of 2030. One of them is based on current policy frameworks and assumptions concerning future 1.3 Objective 5 policies, taking into account recent levels of costs achieved. The other is based on what the industry could do with positive governmental responses to cost reductions and by overcoming barriers to deployment. A screening process and a detailed economic analysis is performed, resulting in LCoE-maps for the Baltic Sea, the North Sea and the Atlantic. Though, this map does not present a detailed LCoE distribution across Danish waters, as it is an Europe-wide study. The LCoE development for offshore wind up until 2030 in the central and southern North Sea is studied in the WINDSPEED project [18] in 2011. It consists of a detailed screening process and a methodology for an integrated assessment of how various constraints - spatial, policy, growth, grid and market integration - impact the deployment potentials for offshore wind energy. However, only a rough estimate on LCoE across the North Sea is given, only presenting high and low costs for offshore wind energy without exact values. The study undertaken by Bogdanov et al. in 2016 [19] provides an image showing the LCoE for offshore wind in 2030 throughout the entire European and Eurasian Sea. Though, no details on how this was obtained are stated, as the paper focuses on the integration and cost of offshore wind energy in the Eurasian Sea for consumption in Europe.

Apart from the above named studies - focusing on the Danish Sea alone or includ- ing it fully or partly through analysis carried out in the European Sea - the LCoE potential for offshore wind has been examined for other countries by means ofLCoE- maps, such as the United Kingdom in [20], Chile in [21] and several for the United States in [22], [23] and [24].

However, no LCoE study for the entire Danish Sea, consisting of a combined detailed cost model and a visualized LCoE distribution for offshore was found. In addition, the majority of the above named studies were developed before the end of 2016, when the most significant drop in history of offshore wind energy costs occurred - when published their winning bid for the offshore wind farm Kriegers Flak of 49.9e/MWh [25]. This was more than a 50% reduction compared to their bid for Horns Rev 3 in the beginning of 2015. Studies carried out before that time are thus not representing today’s costs related to offshore wind energy projects.

1.3 Objective

The objective of this study is to discuss the following questions:

1. Where in the Danish Sea is it most profitable to erect offshore wind farms considering wind resources, water depth and distance to shore?

2. How will potential future development of wind energy technology and the asso- ciated costs influence the Levelized Cost of Energy (LCoE)? 6 1 Introduction

3. Will future Danish offshore wind energy projects be able to operate without subsidies?

In order to discuss the above mentioned questions, a tool to estimate the economic potential of offshore wind energy projects in the Danish Sea is developed bytaking wind resources, water depth and distance to shore into account. The tool enables a comparison of LCoE of offshore wind projects across the Danish Sea in a fastand convenient manner through a graphical visualization. To obtain a representative overview of the most economically attractive areas for offshore wind energy, the user can easily adapt the input parameters such as wind turbine type (capacity), wind farm capacity, turbine spacing within the farm layout as well as economical input parameters. In addition, the model presents an instrument to estimate future devel- opment of LCoE until 2030, considering cost development scenarios and technological improvements such as wind turbine capacity, lifetime and wind farm capacity. Also, this study provides an analysis of energy production potential at selected sites by the Danish Energy Agency (DEA) [1], resulting in a ranking of the sites according to the their LCoE. Finally, the LCoE output of the developed tool is compared to the electricity price prediction until 2030 provided by the DEA [3], leading to a discussion about a subsidy-free future for offshore wind projects in the Danish Sea.

1.4 Content

This Thesis consists of seven chapters. The present chapter 1 provides an introduction to offshore wind energy and defines the motivation and objective of this study. Chap- ter 2 presents the background for this Master Thesis project. It explains briefly the most relevant concepts of the offshore wind turbine technology. Chapter 3 outlines the methodology used by describing the theoretical basis of the LCoE computation as well as the setup of the LCoE-model. Chapter 4 presents all relevant input data for the LCoE-model. Hereby focusing on the technical and economical input parameters. Chapter 5 analyzes the obtained results, starting out with the Cost of Energy (CoE) and continuing with the LCoE. Furthermore, a ranking of LCoE of the pre-selected sites is performed. Finally, a comparison of future electricity prices and the obtained LCoE for offshore wind energy projects is introduced, resulting in a discussion ofa subsidy-free future for offshore wind energy in Denmark. Chapter 6 presents thefinal thoughts and most relevant information one should extract from the work carried out along this Master Thesis. Chapter 7 contains the outlook. It points out the limita- tions of this project and describes several options to further develop the presented methodology. CHAPTER 2 Wind turbine theory

This chapter provides a short introduction to the most important concepts in the wind turbine technology, focusing on offshore wind energy generation. Multiple wind turbines are usually gathered and connected in wind farms, which are installed both onshore and offshore. A typical offshore wind farm setup is shown in figure 2.1, including the substructure and the electrical interconnection between turbines and substation.

Rotor diameter

Rotor blade

Nacelle

Hub height

Tower

Substation

Transition piece J-Tube Substructure Water surface

Monopile

Mudline Inter-array cable Transmission cable

Figure 2.1: Basic setup of an offshore wind farm including the offshore substation. The turbines are installed on substructures consisting of monopile and transition piece.

The entire substructure consists of the monopile and the transition piece (TP) - the connecting part between turbine tower and monopile structure. The monopile is a cylindrical steel tube that is piled into the seabed. According to [26] monopiles have been installed up to water depths slightly above 30 m, while [27] mentions water depths up to 40 m or deeper. However, also other types of substructures are used such as gravity, jacket and tripod structures. Additionally, floating substructures are 8 2 Wind turbine theory under development. The substructure type choice depends mainly on local sea bed conditions, water depth and sea states. Today, monopiles and jackets are the most common substructure types. Around 80% of all substructures installed up to now are monopiles [26]. For that reason, the monopile is the only substructure type considered in this study.

Nowadays the three-bladed vertical axis wind turbine is the industry standard. The rotor is coupled to the generator, either directly if it is a direct drive turbine, or through a main shaft and a gearbox that speeds up the rotation. The electrical power produced by the generator through rotation is brought down by the internal turbine cables and further down through the J-Tube that is mounted on the outside of the substructure. From here the inter-array cables are connecting the turbines within the farm and finally leading the collected power to the substation. The transmission to land proceeds from here with a submarine cable that is buried in the seabed [28].

Equation 2.1 determines the power P produced by the wind turbine:

1 P = · ρ · A · v3 · c (λ, θ) (2.1) 2 p where ρ is the air density, A is the swept rotor area, v the wind speed and cp the power coefficient, which is a measure representing a wind turbine’s efficiency asaratioof the power produced by the turbine and the power in the wind flow at a specific wind speed. The power coefficient is dependent on the tip speed ratio λ of the rotor and the pitch angle of the blades θ. The tip speed ratio is a function of rotor radius R, the rotational speed ω of the rotor and the wind speed v:

ω · R λ = (2.2) v

The blades of the rotor are pitched by a certain pitch angle θ to prevent the generator from producing more than rated power when exceeding the so-called rated wind speed. Likewise the thrust force T acting on the turbine rotor can be expressed as:

1 T = · ρ · A · v2 · c (λ, θ) (2.3) 2 t where ρ is the air density, A is the swept rotor area, v the wind speed and ct the thrust coefficient which is a function of the tip speed ratio λ and the pitch angle θ. The power production and the thrust force are thus changing with varying wind speeds. The relation between power output and wind speed is commonly represented by the so-called power curve of a wind turbine. As an example the power and thrust coefficient curve of the Siemens wind turbine SWT-3.6-120 [29] are shown in the following figure: 2.1 Utilized power curves 9

Figure 2.2: Power curve (left y-axis scale) and thrust coefficient curve (right y-axis scale) of the Siemens wind turbine SWT-3.6-120 [29] with rated power of 3.6 MW and a rotor diameter of 120 m. Air density ρ equals 1.225 kg·m−3. (WAsP input file).

The power curve is defined by a cut-in wind speed determining the wind speed where the turbine starts operating (in the example 4 ms−1), a rated wind speed where rated power is produced (rated power of 3.6 MW produced at 14 ms−1) and the cut-out wind speed defining the maximum wind speed under which the turbine can operate (25 ms−1). It has to be noticed that the power production of a wind turbine is influenced by the temperature and thus by the air density. For the sakeof simplicity this dependency is neglected throughout this study.

2.1 Utilized power curves

It is expected that the turbine capacity during the coming decade will increase im- mensely, as the wind turbine market already these days advertises turbines up to 12 MW (General Electric (GE)) [30] to be available in the next few years. Considering the three planned offshore wind farms, it is assumed that the first farm will stillbe equipped with the 8 MW turbine from Vestas, whereas the second farm is designed with the 10 MW DTU reference turbine and the third with a 12 MW turbine sim- ilar to the Haliade-X manufactured by GE [30]. The power curve of the DTU 10 MW reference turbine is reproduced by the partial load efficiency of the drive train consisting of a 2 stage gearbox combined with a medium speed permanent magnet generator. The general information of the 10 MW turbine is extracted from [31] and reproduced by the INNWIND report [32]. As no power curve for a 12 MW turbine is available yet, the DTU 10 MW turbine’s power curve is scaled up by the ratio Aratio 10 2 Wind turbine theory of the swept area of the two rotors, assuming a rotor diameter of the 12 MW turbine of 220 m (similar to the GE Haliade-X turbine). Taking a look at equation 2.1 that determines the power generated by a wind turbine, it is assumed that only the rotor area A changes when up-scaling a turbine. Thus, the scaling factor Aratio becomes:

( )2 ( ) ( ) A π dR,12MW d 2 220 m 2 A = 12MW = ( 2 ) = R,12MW = = 1.528 (2.4) ratio d 2 A10MW R,10MW dR,10MW 178 m π 2 All power curves used in this study are illustrated in figure 2.3.

Figure 2.3: Power curves for the Siemens wind turbine (SWT-3.6MW-120) [29] with rated power of 3.6 MW and a rotor diameter of 120 m, the Vestas (V164-8.0MW) [33] with rated power of 8 MW and rotor diameter 164 m, the DTU reference turbine (DTU 10MW-RWT) [31] with 10 MW rated power and 178 m rotor diameter and a turbine with rated power of 12 MW and a rotor diameter of 220 m (up-scaled from DTU 10 WM turbine). CHAPTER 3 Methodology

In this chapter the concept of Levelized Cost of Energy is introduced as a method to compare the expected energy production cost of different energy producing units. It is explained how the concept is adapted to compare different offshore wind farms. All assumptions and limitations used are presented as well. Finally, a methodology for mapping the expected LCoE for offshore wind farms positioned consecutively across the Danish Sea is introduced.

3.1 Levelized Cost of Energy (LCoE)

The Levelized Cost of Energy (LCoE) is a measure that enables the comparison of cost of different energy generation methods and is given as a price per unitof generated energy (here e/MWh). The LCoE is not only a measure of attractiveness of investment, but also represents the average minimum price at which the produced electricity needs to be sold in order to pay back the investment and operating expenses of the project. According to [32], the LCoE is calculated as follows: ∑ LT CAP EXt+OPEXt t=0 t LCoE = ∑ (1+r) (3.1) LT Et t=0 (1+r)t with CAP EXt capital expenditures at time t, OPEXt operating expenditures at time t, Et energy production at time t, LT lifetime and r discount rate. The discount rate is used to determine a present value of future cash flows of a project and is composed of inflation and risk. This can be expressed as a summation oftheLCoE for CAPEX and OPEX:

LCoE = LCoECAP EX + LCoEOPEX . (3.2) Assuming a constant average annual energy production (AEP) throughout the entire lifetime, the LCoECAP EX becomes: C LCoE = 0 (3.3) CAP EX a · AEP · LT where a is the levelizing factor and C0 is the investment cost in year 0, which in this study is composed of investment for the wind turbine, the substructure, installa- tion, electrical infrastructure and the project planning and development costs. The 12 3 Methodology levelizing factor a becomes: [ ( ) ] 1 1 + r 1 LT +1 a = · 1 − . (3.4) LT r 1 + r The operating expenditures cover the operation and maintenance cost (O&M) for the wind farm which can be separated into the daily operation and the planned and unplanned maintenance as well insurances and condition monitoring. Usually these costs vary from year to year, mainly depending on the extent of the maintenance complexity (e.g. high cost in a year of failing gearbox). However, this study uses a simplified approach. The OPEX is averaged over the entire lifetime, thus assuminga constant cost per year. Thereby the contribution of the OPEX on the LCoE can be expressed as: C LCoE = OPEX (3.5) OPEX AEP which is a constant cost per MWh to be added to the LCoECAP EX . It has to be noted that it is presumed that the given OPEX per MWh includes all losses related to the AEP. [32]

3.2 Assumptions and simplifications

In order to provide a tool that enables the comparison and visualization of the LCoE of offshore wind at various sites in Danish waters, a model is developed thatcom- putes the LCoE for each pixel on a given map, using the above described method (section 3.1). To limit the complexity of this LCoE-model, the following preliminary assumptions and limitations are applied:

• The analyzed time frame ranges from 2008-2030. • The wind farm layout is chosen to be rectangular and with a turbines spacing within the wind farm of 7×rotor diameter. • The water depth and wind resources for each pixel are averaged for the wind farm area and assigned to the current pixel. • The wind resources are combined for all wind direction sectors by weighting the wind resource data according to their sector occurrence frequency. • The minimum distance to shore is 10 km. • The maximum water depth is limited to 50 m since this is considered to be the maximum depth at which a monopile should be installed. • The transmission cable length is computed directly from the substation to one of the selected onshore grid connection points. Thus, no land cable cost is taken into account. 3.3 Model setup 13

• No decommissioning cost is considered.

• No taxes are taken into account when computing LCoE.

3.3 Model setup

The LCoE-model is based on a pixel map with a resolution of 500 × 500 m covering the Danish Sea which can be seen in figure 3.1. The black dashed line in the figure represents the boundary of Danish waters. As it can be seen, not all of the Danish Sea is covered by the map, as the boundaries of the map are chosen according to the limitations of the available wind resource data. However only small parts compared to the entire area of the Danish Sea are missing at the eastern and western end of the map. The LCoE for offshore wind in each pixel across the whole map is computed using equation 3.1. The methodology applied on each pixel is illustrated exemplary in figure 3.1 and explained in the following:

Figure 3.1: Model setup, including LCoE-map boundaries and border of the Danish Sea (black dashed line) (left) and exemplary wind farm modelling around one pixel (right). UTM Zone 32N projection.

The current pixel under computation is considered to be the mid point of the wind farm. A rectangular wind farm layout around it is assumed (see figure 3.1). Across the area of this wind farm, the average water depth is determined as well as the average annual energy production (AEP), considering a specific wind turbine type. Furthermore, the offshore substation is modeled to be placed 500 m from the turbine arrays, halfway between the rows. The length of the submarine transmission cable is then determined as the direct distance from the offshore substation to the closest onshore grid connection point. All economical input data is determined for the year 14 3 Methodology of interest. Finally, the LCoE can be obtained for this wind farm and the value is assigned to the pixel under computation. Thereafter, the model moves to the next pixel and computes the LCoE following the same procedure. This is done until each pixel contains one LCoE value. This matrix can subsequently be illustrated by means of a color map. All maps in this study are presented in UTM coordinate projection, Zone 32N. As the LCoE for the pixel under computation is determined through the wind farm area around it, no LCoE calculation can be performed at the outer boundary of the map. This area, corresponding to half of the wind farm width or length is left blank and will not be shown in this report. The considered technical and economical parameters for the LCoE computation are shown in the following table:

Table 3.1: Input parameters: Technical and economical aspects.

Technical aspects Economical aspects Wind resources CAPEX Water depth Wind turbine type Distance to shore Substructure Wind turbine capacity Turbine and foundation installation Wind farm capacity Electrical equipment: Wind farm losses Inter-array cable Transmission cable Offshore substation & onshore plant Planning and development OPEX Fixed & variable O&M OTHER Discount rate Life time

Some of the technical and economical aspects reveal dependencies, e.g. the sub- structure cost is a function of water depth or the transmission cable expenses are directly linked to the distance to shore. Further dependencies and details regarding the input data are outlined in the following chapter 4. The economical and technical lifetime of the wind farm is set to 25 years before 2020 (final investment decision FID) and 30 years after 2020 up until 2030 (FID), as the DEA requires [2]. The FID is set when the project execution phase begins, after all contracts and permits are in place. CHAPTER 4 Input data

This chapter assesses all input data used in the LCOE-model, starting with the tech- nical aspects such as wind resources and water depth and continuing with the eco- nomical parameters, elaborating capital expenditures (CAPEX) and operational ex- penditures (OPEX) in detail.

4.1 Wind resources and Annual Energy Production (AEP)

The wind climate is commonly described by the Weibull probability density function, which is determined by the A- and k-parameters - scale and shape parameters, respec- tively. The Weibull probability density function f can be calculated using equation 4.1: ( ) ( ) k k−1 − v k v A f(v) = · · e (4.1) A A with v wind speed, k Weibull shape parameter and A Weibull scale parameter. The Global Wind Atlas (GWA) [34] provides this data around the globe, however only for up to 30 km off the coast. Therefore, the GWA-team was contacted andD. Heathfield provided data (A- and k-parameters) for nearly the entire Danish Sea1. The data has a resolution of 500×500 m and is given for heights of 80, 100 and 150 m above sea level, selected according to the hub height of the turbine. To describe the overall wind climate, the A- and k-values are combined for all wind direction sectors, weighted according to their sector occurrence frequency. For the purpose of this study this data is sufficient, since the terrain offshore has no obstacles or significant changes in surface roughness.

1Wind data provided by Duncan Heathfield from World in a Box (www.worldinabox.eu), consul- tant to DTU Wind Energy on the Global Wind Atlas (www.globalwindatlas.info). D. Heathfield provided the following information about the generation of the wind data: Climate data was obtained from the Global Wind Atlas 2.0. Roughness data with 300 m resolution maps was used in which roughness lengths were interpreted from the “CCI” data set using a class to roughness table agreed by Risø scientists in 2018-11 (ESACCI 2015 land cover (https://www.esa-landcover-cci.org/) + translation table (D. Heathfield, pers. comm.)). Elevation data with 150 m resolution maps de- rived from viewfinder data was used (http://viewfinderpanoramas.org/). Both the roughness and elevation maps will be used in the coming release of the Global Wind Atlas version 3.0. 16 4 Input data

The mean wind speed across the Danish Sea is shown in figure 4.1 and was deter- mined by equation 4.2 [35]: ∫ v¯ = f(v) · v dv (4.2)

Figure 4.1: Mean wind speed map of the Danish Sea, including boundary of the Danish Sea (black dashed line) and selected sites by the DEA [1] (purple line).

The AEP for a specific wind turbine with the power curve P (v) can be determined by equation 4.4: ∫ AEP = f(v) · P (v) dv [MWh] (4.3)

In order to clarify this relation, an exemplary Weibull distribution with k = 2 and A = 10 ms−1 is shown in figure 4.2 together with the power curve of the Vestas V164-8.0MW wind turbine [33]: 4.1 Wind resources and Annual Energy Production (AEP) 17

8 Power curve V164-8.0MW 0.08 7 Weibull pdf with k=2 & A=10 6 0.06 5

4 0.04 3 Power [MW] Power Probability [-] Probability 2 0.02 1

0 0.00 0 5 10 15 20 25 30 35 Wind speed [ms−]

Figure 4.2: Weibull probability density function (red) with k = 2 and A = 10 ms−1. Power curve of the Vestas wind turbine V164-8.0MW [33] (blue).

In this study only discrete values for the Weibull distribution and the power curve are available. Thus, the AEP is calculated by the cumulative Weibull distribution as follows: ∑N Pi−1 + Pi AEP = N · [f(v ) − f(v − )] · [MWh] (4.4) h i i 1 2 i=1 with Nh number of hours in one year (8760 h), N number of wind speed bins (v = −1 −1 1 − 30 ms with step size 1 ms ), vi wind speed in bin i and Pi power generation in bin i. It has to be noticed that the AEP is reduced due to losses within the wind farm. A detailed description of the losses is provided in section 4.6. Along this study, the AEP before taking into account losses is indicated by the term Gross AEP. The AEP after losses is called Net AEP.

4.1.1 Validation of wind data A validation of the data provided by D. Heathfield against the original data published on the Global Wind Atlas (GWA) web [34] was performed. The Weibull A- and k- values seemed to correspond very well for onshore sites. However, a comparison at offshore sites revealed major differences. For example a site in the North Sea ((436670, 6139759) UTM Zone 32) was analyzed. The provided wind data gave A = 8.73 ms−1 and k = 2.10 for the weighted Weibull parameters at 100 m height, while the GWA resulted in A = 10 ms−1 and k = 2.29. A validation of the provided wind data was subsequently performed by the GWA team. A latent error was found that was related to the interpretation of the roughness for water in the model. In WAsP2 12.0, some updates within the core code were made which changed the way water roughness was determined. The latent problem became actual when combining the new WAsP

2Wind Atlas Analysis and Application Program provided by DTU Wind Energy [36]. 18 4 Input data core code with old data, resulting in discrepancies for offshore data. The error was resolved and new data provided, which indeed was identical with the data one can find on the GWA web. This new data is used throughout the entire study.

4.2 Water depth

The water depth data used in this study is provided by The European Marine Ob- servation and Data Network (EMODnet) [37] and is based on a EMODnet Digital Terrain Model generated for European Sea regions from selected bathymetric survey data sets and Satellite Derive Bathymetry data products [37]. Gaps with no data coverage are completed by integrating digital bathymetry. The data is provided for 1 free and can be obtained with a resolution of 16 arc minute (equals approximately 115 m). This data is then fitted to the same boundary as the wind resource dataand averaged in order to provide the same pixel size as the wind resource data. Water depths deeper than 50 m are excluded from the model since only monopile substruc- tures are considered in this study. Also, every pixel closer than 10 km to shore is not considered as the upcoming wind farms are planned to be far from shore (DEA states in [6] a distance to shore for near-shore wind farms of 4-10 km). The water depth for the area under study can be seen in figure 4.3:

Figure 4.3: Water depth of the Danish Sea, excluded depth >50 m (marked dark grey) and distance to shore <10 km (marked light grey). 4.3 Capital expenditures (CAPEX) 19

4.3 Capital expenditures (CAPEX)

This section describes all input data related to the CAPEX. This data research turned out to be rather difficult. Barely no economic data is publicly available due tostrate- gic management and competitiveness among the fast growing offshore wind industry. Another reason for the lack of available cost documentation is the fact that the off- shore wind energy industry is a relatively young renewable energy sector. However, literature from different years and authors was reviewed and the available datais used to estimate the development of various cost factors. The data is obtained from projects that have been carried out in the past or from developed models. It has to be kept in mind that the costs are subject to uncertainty since they are very project specific, depending on site conditions such as weather, location and accessibility, com- petitiveness and timing. If data is obtained through models, most likely assumptions and simplifications have been applied. In order to estimate future cost levelsfor offshore wind energy, a baseline cost scenario has been developed as well asahigh and a low cost scenario based on the predicted cost reductions and uncertainties in [6]. All main factors affecting the total CAPEX of an offshore wind farm –turbine, substructure and electrical infrastructure costs - are expected to decrease significantly in the future as a consequence of research and development efforts, industrialization of the sector and economies of scale as well as growing competitiveness and supply chain and service improvements [6]. Only data from 2008 and on-wards is used in the model to avoid impact through the economic crisis in 2008. All costs in this report are reproduced by using the appropriate exchange rates and inflation and are given as e2017. The expenses applied in the model are considered in the final investment decision year (FID).

4.3.1 Wind turbine cost The capital expenditure of the wind turbine is usually given as price per MW. A price development over the past years for wind turbines (assumed onshore and offshore, as nothing else is stated) is given by [38]. A linear regression curve (black line) is added to the figure 4.4 as per best guess, to estimate the turbine expenses used in this study. It is placed on the rather high end of costs, since offshore turbines are associated with higher expenses than onshore turbines in order to withstand the rough offshore conditions. 20 4 Input data

Figure 4.4: Wind turbine price trends from 1997-2017 reproduced from [38] with permission from publisher IRENA, added a linear regression curve manually from 2008-2017 (black line), used in this study.

It can be seen that turbine prices have been increasing from around 2002 until 2008-2009. This can mainly be explained by three factors [38]. Firstly, increasing commodity and labour costs in the years before 2009 resulted in higher turbine prices. Secondly, demand for wind turbines exceeded supply and lastly the manufacturers introduced larger, more powerful turbines, translating into higher prices. The decrease of turbine prices after 2009 is driven by higher competition between manufacturers, supply chain improvements, manufacturing process improvements as well as economy of scale and falling commodity prices [38]. In the present study no variations in costs due to specific offshore site conditions are made. In reality though, the cost may change from site to site due to the risk as- sociated with certain environmental conditions or implementation of site and project specific technologies. It is assumed that future wind turbine cost will continue to drop, though not with the same steep slope observed through the past decade. This assumption is based on the limiting factors such as commodity prices and slowly exploited technology development. To be able to predict the cost reductions in the future, the approach of learning rates is introduced. Learning rates (LR) represent a certain percentage of cost reduction for a technology for every doubling of installed capacity, caused by increased experience, expertise and industrial efficiency associated with deployment and ongoing research and development. In [39] a study was performed by the Danish Energy Agency (DEA) for offshore wind energy in Europe and was compared toother 4.3 Capital expenditures (CAPEX) 21 studies. The study determines a learning rate for offshore wind energy CAPEX of 10% up to 2030. Considering the predicted growth for capacity given by IRENA in 2016 [5] (Introduction figure 1.1), a CAPEX cost reduction of approximately 16% from 2017 to 2030 is determined and applied for the turbine cost. The overall turbine cost function over time is shown in figure 4.5, including a band of uncertainty. The uncertainties are based on the report developed by the DEA [6], resulting in an uncertainty of ±5% in 2020 and ±10% in 2030. Though, it is assumed that the cost will not exceed the current cost level.

Figure 4.5: Wind turbine cost development from 2008-2030. 2008-2017 obtained through [38], 2017-2030 learning rate (LR) of 10% and therefore a cost reduction of 16% is applied.

4.3.2 Substructure cost

In this report it is assumed that monopiles will be installed up until 50 m in the future. For that reason, only monopile substructures have been considered and no LCoE is calculated at water depths deeper than 50 m. The cost of monopiles increases with deeper waters, since larger structures and thus greater amounts of steel is needed. A cost model for substructures depending on water depth was conducted by the Department of Wind Energy at DTU in 2015 [27], which is used as input for this model. It is based on a mix of data from Rambøll, EON, Douglas-Westwood and BVG Associates. It covers the cost of the monopile itself (for 2-3.6 MW wind turbines), the transition piece (TP) and the secondary steel (J-tube, boat-landing platform, corrosion protection etc.). The reproduced substructure cost function is illustrated in the following figure 4.6: 22 4 Input data

Substructure cost function 2015

1.2 Rambøll EON

/MW] Douglas-Westwood 1.0 BVG

2017 DTU fit € 0.8

0.6

0.4

0.2 Substructure CostSubstructure[M 10 20 30 40 50 Water Depth [m]

Figure 4.6: Substructure cost as function of water depth, developed by DTU - De- partment of Wind Energy in 2015 [27], reproduced for this study.

In order to estimate the cost development in the past, cost specifications for monopiles at 20 m water depth have been extracted from various studies and a linear regression is applied to this data (see figure 4.7). The development shows a reduction in substructure cost. To verify this development, the obtained substructure cost is compared to the steel price development, since the monopile and TP are mainly pro- duced out of steel. An overall decline of steel prices is observed from 2008-2016 [40] corresponding to the latter mentioned reduction in substructure cost. It is assumed that the cost of the substructure will continue to decrease in the future. One reason for this assumption is an expected reduction of substructure cost per MW due to increasing turbine capacity. It has to be noticed that the provided cost function in figure 4.6 is developed for turbine capacities ranging from 2-3.6 MW. For larger turbines, also the monopile needs to be designed larger and the overall cost will increase. Though, the cost per MW will decrease, as less material is needed relative to the capacity increment. Another reason for the assumption that substructure costs will decrease is the ongoing improvement and automation of the manufacturing process. However, the cost reduction is assumed to have a less steep slope compared to the previous years, since there are limiting factors such as commodity prices. Because no predictions for future monopile price levels were found, the learning rate of 10% for the CAPEX - as discussed in the previous section 4.3.1 - is introduced for the substructure cost development as well, meaning an overall price reduction of 16% from 2017-2030 is applied. The entire assumed price development from 2008- 2030, including the same uncertainty assumptions as for the turbine investment cost (±5% in 2020 and ±10% in 2030), is shown in the following figure 4.7: 4.3 Capital expenditures (CAPEX) 23

Figure 4.7: Substructure cost development at 20 m water depth from 2008-2030. 2008-2017 linear regression obtained through past data from Kooijman [41], Douglas- Westwood [42], Danish Energy Agency [43] and NREL [44]. 2017-2030 learning rate (LR) of 10%, thus cost reduction of 16% applied.

4.3.3 Turbine and substructure installation cost The installation cost is mainly based on the expenses associated with the turbine and substructure installation vessel. Theoretically the installation cost is influenced by the distance to shore, as the cost for installation vessels increases with longer transportation time. Also, deeper water requires longer and heavier substructures, which are more difficult to handle. In addition, the seabed condition and water depth affect the vessel cost, since the monopile driving procedure extends. However, [45] has performed a study taking information from 87 offshore wind farms installed from 2000-2017 into account and concludes that neither water depth nor distance to shore have a very significant impact on substructure installation time. Thus, it is decided not to include distance to shore nor water depth when calculating the installation vessel cost. [45] as well investigates the average installation time for substructures and turbines based on the same 87 wind farms. An installation time for monopiles (MP) plus transition piece (TP) is determined to amount to 0.5 days per MW for projects starting construction work between 2007-2014. In the same manner the turbine installation time amounts to 0.6 days per MW. Installation costs are generally broken down to harbour and storage costs, fixed (mobilization and demobilization) vessel costs as well as vessel spread day rates. For the scope of this synthetic analysis, the installation costs are assumed to be only related to vessel spread day rates. The vessel installation market is not transparent due to the reduced number of competitors. Vessel spread costs depend among other variables on the spread selected (jack-up or floating installation vessels, barges, float- ing cranes) and market competition. As documented in [46] and [47] vessel day rates 24 4 Input data vary significantly, therefore a day rate of 150, 000e is used throughout this report. The installation cost details are specified in table 4.1:

Table 4.1: Turbine and substructure installation time and installation vessel cost.

days days e2017 Turbine inst. time [ MW ] MP+TP inst. time [ MW ] Vessel Cost [ day ] 0.5 0.6 150,000

4.3.4 Electrical equipment and infrastructure cost This chapter deals with the electrical infrastructure of offshore wind farms and the associated costs. Generally the low voltage output of each wind turbine (usually 690V) is stepped up by the turbine’s internal transformer to medium voltage (33kV or 66kV). The turbines are connected by this medium voltage network (inter-array cables), which is leading the power to the substation. From here the transmission to land proceeds with either a high voltage alternating current (HVAC) or high voltage direct current (HVDC) sub sea transmission cable that is buried in the seabed. The onshore substation might need an additional transformer to step up the voltage to match the transmission grid voltage [28]. The chapter is divided into two sections - the inter-array connection and the transmission including the sub sea cable and the offshore and onshore substations.

Inter-array connection The voltage level for inter-array cables in this study is set to 33 kV, which is commonly used in today’s wind farms. Noteworthy is however that the trend for future inter- array cables will shift to a voltage level of 66 kV as turbine generator capacities are increasing [48]. This trend is also reflected by currently available turbines. The Vestas turbines V164-8.0MW and V164-9.5MW are available for both 33 and 66 kV, while the V164-10.0MW is available for 66 kV only [33]. A higher voltage level for the array cabling will translate into lower current for a given power, which will allow the transport of greater power along the cable with the same cross sectional area. This means that more turbines can be connected to one cable and thus decrease the number of cables needed. However at the same time, shifting to a 66 kV cable will result in higher expenses related to electrical equipment for substation and wind turbine transformer [48]. In this study all computations are performed using a 33 kV inter-array cable due to a lack of available data for 66 kV cables and to be able to validate the model on past wind farm data. The wind farm layout is based on a rectangular turbine arrangement, spaced with a distance of 7×rotor diameter in both the crosswind and downwind direction. No wind farm layout nor cable layout optimization is performed in this study. [49] provides offshore inter-array cable costs for a voltage level of33kV 4.3 Capital expenditures (CAPEX) 25 alternating current (AC). The costs are given by two companies for different copper 2 cross sectional areas ACu of the cable up to 800mm and a corresponding maximal steady state current Issn of 900 A. An extrapolation of the steady state current and the price is performed in order to be able to utilize a bigger sized cable of up to 1050 A. Consequently more turbines can be connected in one row. Since the cable distances in between the turbines are short (around 1 km for a V164 turbine with a spacing of 7×rotor diameter) cable losses are neglected. The key parameters of the chosen inter-array cable are shown in the following table 4.2:

Table 4.2: Inter-array cable key parameters with Vinter−array inter-array cable voltage, ACu cross-section of cable, Issn [A] steady state current and Cinter−array cost per length of the inter-array cable.

2 −1 Vinter−array [kV] ACu [mm ] Issn [A] Cinter−array [e2017·m ] 33 1000 1050 615

This choice of cable is determining the wind farm layout (number of rows and columns) in this study, as only a certain number of turbines can be connected in a row to one cable of that size. The number of turbines NT in a row is calculated as following: √ 3 · Vinter−array · Issn NT = (4.5) Prated,turbine where Prated,turbine is the rated power of the turbine, Vinter−array the voltage and Issn is the steady state current of the inter-array cable. An exemplary wind farm layout is shown in figure 4.8:

Figure 4.8: Exemplary wind farm inter-array cable layout including substation and transmission cable. 26 4 Input data

The substation is modeled to be placed 500 m from the turbine arrays, halfway between the rows, closest to the next grid connection point. It has to be mentioned that in real life different cable sizes within a row can be chosen to lower the overall expenses. To keep this model simple and to decrease computational time, it is decided to only conduct the computations with one cable size. However, no cable installa- tion costs are included since the cables and the associated costs are conservatively estimated.

Transmission To connect an offshore wind farm to the electrical grid a strong point of common coupling (PCC) is needed. For this purpose suitable 400 kV connection points in the Danish grid [50] have been selected considering current or future utilization, such as connection points for offshore wind farms or transmission between different grids (sketched in Appendix figure A.2):

Table 4.3: Suitable 400 kV onshore points of common coupling (PCC).

Region of Substation Current utilization Denmark Jutland Ferslev HVDC transmission Denmark-Sweden. Idomlund Connection point of Vesterhav Nord offshore wind farm. Trige Connection point of Anholt offshore wind farm. Endrup Connection point of Horns Rev 1-3 offshore wind farm. Fyn Kingstrup Connection to Jutland. Sealand Herslev Connection between DK1 and DK2 transmission. Bjæverskov Connection point of Kriegers Flak offshore wind farm.

The distance from the offshore wind farm substation to the onshore PCC iscom- puted directly as a straight line, meaning that no onshore cable costs are taken into account. Considering the transmission cable and the offshore and onshore substation, a distinction between high voltage direct current (HVDC) and high voltage alternating current (HVAC) is made in terms of power transmission capability and expenses. In direct current (DC) the electric current only flows in one direction whereas the electric current in alternating current (AC) changes direction periodically. Also, the voltage in AC circuits periodically reverses since the current changes direction. The overall cost for the transmission system can be separated into cable costs and electrical plant costs (offshore and onshore substation) [51]. The onshore plant is related to costs needed to step-up the voltage of the transmission system to 400 kV of the PCC. HVDC has the advantage that no electrical losses due to charging current (current associated with the capacitance of a cable) occur during transmis- sion. However, at the same time it has to be invested in costly substation supplies 4.3 Capital expenditures (CAPEX) 27 arising from power converter equipment. This is caused by the need for an AC-DC conversion in the offshore substation and a DC-AC conversion on land to match the AC grid. HVAC has relatively inexpensive substation costs, but due to reactive power transmission, the AC cable is exposed to higher electrical losses which are increasing with distance [51]. Therefore, HVAC is usually preferred for shorter distances while HVDC for longer distances from offshore substation to land. The HVAC and HVDC transmission system setups are illustrated in figure 4.9:

Figure 4.9: Upper: high voltage alternating current (HVAC) transmission system setup, Lower: high voltage direct current (HVDC) transmission system setup [51].

The approach of comparing transmission capability and costs of HVDC and HVAC used in this model, is based on the study performed in [51].

HVAC

The costs of the AC offshore substation CSHV AC and the onshore plant COHV AC are given by equation 4.6 and 4.7, respectively:

CSHV AC = 6.191Me + 0.056Me/MVA · ST (4.6)

0.930 COHV AC = 0.0325Me/MVA · ST (4.7) with ST as transmission power rating in [MVA], taking into account offshore trans- former platform fixed and variable costs plus transformer and converter number and costs. To estimate the cable costs, the cable transmission capability has to be inves- tigated. The reactive power produced by the capacitive charging current Qc can be expressed by: 2 Qc = VHV AC · 2 · π · f · C · lc (4.8) with VHV AC nominal voltage (here 220 kV), f system frequency of 50 Hz, C capaci- tance per length of the cable and lc length of the cable in [km]. The cable is designed so that the compensation power is evenly distributed on both ends of the cable in order to increase the active power current and thereby the 28 4 Input data

cable transmission capability Pc for a three core AC cable, which is then given by equation 4.9: √ ( ) Q 2 P = S2 − c c c 2 √( ) ( ) (4.9) √ 2 1 2 = 3 · V · I − · V 2 · 2 · π · f · C · l HV AC ssn,HV AC 2 HV AC c with Sc cable power rating, Qc reactive power, Issn,HV AC steady state current, VHV AC HVAC voltage, f system frequency of 50 Hz, C capacitance per length of the cable and lc length of the cable in [km]. [51] provides cable specific parameters for cables with various sizes for different voltages. The 220 kV cable with the greatest steady state current was selected for this study since wind farm capacities tend to increase in the future. The HVAC cable key parameters including the cable cost can be found in the following table 4.4:

Table 4.4: HVAC sub sea cable key parameters with VHV AC HVAC nominal voltage, Issn,HV AC as steady state current, ACu cross-section of cable, C capacitance per length of the cable and Ccable,HV AC HVAC cable cost per length.

2 −1 −1 VHV AC [kV] Issn,HV AC [A] ACu [mm ] C [nF·km ] Ccable,HV AC [e2017·m ] 220 942 1000 177 1240

By equation 4.9 the maximal transmission capability for the cable with a specific length is obtained. In this manner the number of HVAC cables Nc needed for a specific wind farm capacity is found and following the total HVAC transmission cable cost CCHV AC is determined:

CCHV AC = Ccable,HV AC · lc · Nc (4.10) with Ccable,HV AC HVAC cable cost per length, lc length of the cable and Nc number of cables needed. Finally, the entire cost related to the HVAC transmission becomes:

CHV AC,total = CSHV AC + COHV AC + CCHV AC (4.11) with CSHV AC cost of offshore substation, COHV AC cost of onshore substation and CCHV AC transmission cable cost for HVAC.

HVDC

The expenditures for the HVDC offshore substation CSHVDC and onshore station COHVDC are significantly higher than for HVAC and calculated by equation 4.12 and 4.13: CSHVDC = 30.957Me + 0.147 · Me/MVA · ST (4.12) 4.3 Capital expenditures (CAPEX) 29

COHVDC = 0.101Me/MVA · ST (4.13) with ST as transmission power rating in [MVA]. The HVDC cable is not subject to charging current losses. The steady state current for the two polar HVDC cable is given by: ST Issn,HV DC = (4.14) 2 · VHVDC with Issn,HV DC as steady state current of the HVDC cable, ST as transmission power rating and VHVDC as nominal voltage. The voltage here is set to 300 kV. As it can be seen, equation 4.14 is not subject to losses due to the length of the cable. Knowing the steady state current of the wind farm, the price can be determined by usage of the following relation shown in figure 4.10, derived from the provided cost relations in [51] and added a polynomial regression curve used in this study.

HVDC cable cost @ 300 kV

1400 /m] 1200 2017 € [ 1000 HVDC CC 800 Poly fit El. parameters

1200 1400 1600 1800 2000 2200 2400 2600

Issn [A]

Figure 4.10: HVDC cable cost over steady state current with voltage of 300 kV. Added polynomial regression curve to prices given by [51].

Finally, the entire cost related to the HVDC transmission becomes:

CHV DC,total = CSHVDC + COHVDC + CCHVDC (4.15) with CSHVDC cost of offshore substation, COHVDC cost of onshore substation and CCHVDC transmission cable cost for HVDC. The break-even length - known as the length where HVDC becomes cheaper than HVAC - varies between approximately 65 km and 100 km for wind farm capacities from 400-1000 MW in this study. An exemplary graph of HVAC vs. HVDC costs is shown in figure 4.11. The stepwise increase of HVAC cost is explained by the growing number of cables needed, whereas the increase of HVDC cost in this study only depends on the cable length. 30 4 Input data

Transmission cost for a 800 MW wind farm - HVAC vs. HVDC

HVAC HVDC 3000

2500

] 2000 2017 €

1500 Cost[M

1000

500

Break-even length 97 km 0 0 50 100 150 200 250 Cable Length [km]

Figure 4.11: HVAC vs. HVDC electrical equipment cost for an exemplary wind farm with a capacity of 800 MW, including cost for onshore and offshore substation and transmission cable. Break-even length at approximately 97 km.

Development of grid connection costs According to the Danish Energy Agency in [6] the cost associated with the grid connection is expected to decrease in the future. Though, the LR applied to the CAPEX and OPEX is not used here, since some parts of the grid connection are mature while others are not. Therefore a cost reduction of 1% per year is used up until 2020 and 0.75% reduction per year until 2030 [6]. 4.4 Operational expenditures (OPEX) 31

4.4 Operational expenditures (OPEX)

The operational expenditures (OPEX) represent the annual fixed and variable expen- ditures required to operate and maintain a wind farm over its technical lifetime. The operation and maintenance (O&M) cost covers scheduled and unscheduled mainte- nance of turbine components as well as replacement of components. Apart from that, the OPEX also includes insurance, taxes and other costs such as project management and administration, weather forecasting and condition monitoring [52]. Usually the OPEX is influenced by various factors such as O&M vessel expenses, climate conditions (weather windows), distance to shore, more precisely distance to closest port and labour costs. In some cases the water depth has an influence as well, e.g. if a jack-up vessel is needed to replace bigger turbine parts such as blades. Hence, the OPEX is subject to many different factors which are very project specific, also depending on for instance competition on the market or the type of O&M contract signed. Therefore and due to a lack of available data, it is difficult to estimate the OPEX for a specific site. To be able to simply add the OPEX as a cost per MWh produced on top of the LCoE - as explained in section 3.1 - the OPEX prediction provided by the Danish Energy Agency in [6] is fed into the LCoE-model. The DEA assumes in [6] that 75% of the OPEX is covered by fixed costs and 25% by variable costs that vary with production. All OPEX are given as expenses per MW and year for the fixed costs and as expenses per MWh for the variable costs. According to [6] the provided O&M costs are based on the average OPEX for existing offshore wind farms owned by Ørsted and further interviews with the industry. To obtain a price per MWh the OPEX per MW and year is divided by the full load hours of the wind farm (ratio of annual energy production and rated power of the wind farm), assuming the capacity factors given in table 4.5 (provided by [6] as well) and 8760 hours per year. As for the CAPEX, the OPEX development until 2030 is obtained by applying a learning rate of 10%. Also the uncertainties are the same as for the CAPEX.

Table 4.5: Capacity factor prediction of [6] and the resulting OPEX determined by assuming 75% fixed and 25% variable cost and 8760 hours per year.

2015 2020 2030 Capacity factor [%] 50 51 53 OPEX [e/MWh] 18 13 11 32 4 Input data

Resulting in a regression line as shown in figure 4.12:

Figure 4.12: OPEX development. Cost scatter up to 2015 derived from various sources: NREL [53], Ernst & Young [54], Engels [55], INNWIND [32], IEA [56]. OPEX prediction obtained by the Danish Energy Agency and Energinet [6]. Predic- tions for comparison given by BVG Associate [15].

The cost predictions given by BVG Associate [15] are only included as a com- parison to the OPEX provided by the DEA in [6]. As it can be observed, the cost forecasts match quite well. BVG Associate’s predictions are placed slightly higher than DEA’s, however getting closer later on in time.

4.5 Other expenses

A part of the CAPEX is associated to management plus project planning and de- velopment of the wind farm, including the environmental impact assessment (EIA), engineering and permitting. In addition insurances have to be paid as well as civil works. An overall cost of 10% of the CAPEX is assigned to the above mentioned costs, as stated in [6]. Usually these expenses depend on the planning duration and site conditions e.g. seabed characteristics or eventually nature protected areas as well as the overall market condition.

4.6 Wind farm losses

The so-called wake effects amount to the greatest fraction of the overall losses ina wind farm. Wakes are defined as wind velocity deficits behind a turbine, resulting ina reduced power density at the following turbines downstream and thus causing annual 4.6 Wind farm losses 33 energy production (AEP) reductions. In order to roughly estimate the occurring AEP losses due to the wake effects, an analysis is performed in WAsP (Wind Atlas Analysis and Application Program) [36]. WAsP is a software used for , siting and energy yield calculations for wind turbines and wind farms. However, the wake model implemented in WAsP is not designed for the computation of wake losses for large offshore wind farms. For this purpose the software Fuga[57] was developed as part of the WAsP package. Fuga is a linearized CFD model that is used to estimate wakes within large offshore wind farms. The wake losses are estimated by using three wind turbine types: Siemens SWT- 3.6-120, Vestas V164-8.0MW and the DTU reference 10 MW turbine. The 12 MW turbine is not used as no thrust curve is available for that turbine. A wind farm setup with 100 turbines and varying spacing (6, 7 and 8×rotor diameter) is designed for each turbine type. Layouts of 5 × 20 turbines are placed 20 km from shore (closest turbine to land). The columns of 20 turbines are facing west, since the prevailing wind directions are west-northwest to south-southwest counterclockwise. This is illustrated by the wind rose in figure 4.13:

Figure 4.13: Wind rose and Weibull distribution at 100 m height (WAsP derived). Data obtained from the Global Wind Atlas [34] at position UTM Zone 32 (422042, 6206784) placed 20 km west of the Danish coast of Jutland. With Weibull A and k weighted for all sectors, U mean wind speed and P mean power density.

As an example, the layout with 7×rotor diameter for the 8 MW Vestas turbine and the corresponding wake loss plot derived in Fuga are shown in the following figure: 34 4 Input data

(a) Wind farm position and layout (WAsP). (b) Wind farm wake loss plot (Fuga).

Figure 4.14: Wind farm layout and wake losses for a 800 MW farm consisting of 100 Vestas V164-8.0MW turbines placed 20 km west of the Danish coast of Jutland (Fuga derived).

Figure 4.15 showcases the resulting wake losses for all three turbines depending on the spacing between the turbines.

Figure 4.15: Wake losses depending on turbine spacing (6, 7 and 8×rotor diameter (dR)) and turbine type (Siemens SWT-3.6-120, Vestas V164-8.0MW and DTU 10MW RWT). 4.6 Wind farm losses 35

Concluding on the above shown plot, wake losses are varying between approxi- mately 12% and 7% depending on turbine type and distance between the turbines. As the wake losses for a spacing of 7×rotor diameter, which is applied to determine the inter-array cable length, range between 8.4 and 9.5%, an overall wake loss value of 9% is used throughout this study. However, it has to be mentioned that the wake losses generally depend on multiple factors, such as atmospheric boundary layer, wind resources at site and . As no design optimization is performed in this study, the same wake loss of 9% is applied for all sites (pixels on the map) throughout the entire Danish Sea. Other AEP losses are associated with turbine availability which is equivalent to loss of production due to scheduled and unscheduled maintenance services. A value of 97% of turbine availability, corresponding to 3% availability losses is applied [58]. Furthermore, internal electrical losses related to cable and substation losses of 1-2% are stated in [58], therefore 1.5% is assumed in this study. Other losses can be related to turbine performance (power curve adjustments, control losses) or environ- mental losses (blade icing, differences in temperature). However, these losses are not considered here. The overall losses on AEP used in this study are resulting in (1-(0.91 · 0.97 · 0.985) = 13%) and are summarized in table 4.6.

Table 4.6: Wind farm AEP losses.

Type Value [%] Turbine availability losses 3 Wake losses 9 Electrical losses 1.5 Total losses 13 36 CHAPTER 5 Results

This chapter presents the main results of this study. First, the Cost of Energy (CoE) is introduced which is used to estimate the discount rate for offshore wind energy projects. Following, the Levelized Cost of Energy (LCoE) distribution for offshore wind across the Danish Sea is visualized and the LCoE predictions up until 2030 are analyzed. The cost reductions described in chapter 4 and technology improvements are taken into account. Furthermore, a ranking according to LCoE of the pre-selected sites (given by the Danish Energy Agency (DEA) [1]) is presented. Finally, a discus- sion regarding a subsidy-free future for offshore wind energy projects is introduced by comparing electricity price predictions (DEA [3]) and the obtained LCoE development for offshore wind in this study. 38 5 Results

5.1 Cost of Energy (CoE)

The Cost of Energy (CoE) can be evaluated at the end of the lifetime of an energy producing unit by dividing all costs related to installation and operation by the total energy produced. This is equivalent to using the LCoE equation 3.1 in section 3.1 with a discount rate of 0%. The levelizing factor a (equation 3.4) then becomes a = 1. Since the discount rate is always increasing the LCoE one can estimate a lower bound for the LCoE by calculating the CoE first. The equation for the CoE is therefore give by: C CoE = 0 + C (5.1) LT · AEP OPEX with C0 investment in year 0 (capital expenditure (CAPEX)), LT lifetime, AEP annual energy production and COPEX operational expenditure (OPEX) per MWh. The CoE is determined for each 500×500 m pixel throughout the entire map of the Danish Sea (figure 1.2). The methodology explained in chapter 3 is used to assign a CoE value to each pixel. The pixel under computation is considered to be the mid point of the wind farm. A rectangular wind farm layout is assumed around the pixel. Across the area of this wind farm, the average AEP is determined, considering a specific wind turbine type. Applying the AEP losses (section 4.6) the Net AEP is found for that pixel. The mean water depth is determined equally. Furthermore, the offshore substation is modeled to be placed 500 m from the turbine arrays, halfway between the rows. The length of the submarine transmission cable is then determined from the offshore substation to the closest point of common coupling (PCC) tothe grid. All economical input data is determined for the year of final investment decision (FID) and described in chapter 4. To obtain the investment cost C0 in equation 5.1, the costs of the turbine, the substructure, electrical equipment, installation and planning & development have to be obtained. The turbine cost for the year of FID is taken from figure 4.5. Knowing the mean water depth, the substructure cost is obtained through figure 4.6 and 4.7. The inter-array cable cost and the transmission cost (knowing the distance to closest PCC) are determined as described in section 4.3.4. Finally, the turbine and substructure installation cost (section 4.3.3) as well as the planning and development cost of 10% of the CAPEX (section 4.5) is found. The OPEX is obtained in figure 4.12. Knowing all the input parameters in equation 5.1, the CoE can be obtained for this wind farm and the value is assigned to the pixel under computation. This procedure is applied consecutively to each pixel on the map. This obtained CoE-matrix can subsequently be illustrated by means of a color map, which is shown in the following figure 5.1. 5.2 Calibration of discount rate 39

Figure 5.1: Cost of Energy (CoE)-map for the first 800 MW wind farm, tendered in 2019/2020 and assumed year of FID (Final investment decision) in 2021. Input parameters: Vestas V164-8.0MW turbine, lifetime of 30 years, AEP (annual energy production) determined using method in section 4.1, CAPEX (capital expenditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12. Areas closer than 10 km to shore marked light grey, areas with water depth deeper than 50 m marked dark grey, boundary of the Danish Sea indicated through black dashed line and the selected sites by the Danish Energy Agency (DEA) [1] surrounded by purple line.

The CoE-map ranges from slightly under 30 e/MWh to 44 e/MWh. However, the CoE is not representing the true cost, as no time value of money is considered, meaning the cost is not levelized. For that reason, a discount rate has to be applied. Since it is not easy to estimate the discount rate, as it strongly depends on the specific project, a calibration of the discount rate is performed in the coming section.

5.2 Calibration of discount rate

The discount rate, also known as weighted average cost of capital (WACC) [59], is a crucial input when performing economic evaluation. It is used to determine a present value of future cash flows of a project. In private economic analysis the discount rate 40 5 Results is composed of inflation and risk (project risks, political risks etc.) which islinked to uncertainty of projections of the future. It is highly dependent on the business model chosen. Usually there are two sources of capital - equity and debt [59]. For this study a pure equity financing is assumed, so that the discount rate reflects the interest on the entire equity. The discount rate is difficult to estimate, since the risk is very project specific and barely no information is published. However, it is supposed that the discount rate used in the wind energy sector has been falling during the past decade as risk is decreasing with the maturing offshore wind industry. Therefore, the CoE output of the developed model and the LCoE for various ex- isting wind farms or farms under construction provided by the Danish Wind Industry Association (DWIA) [60] are used to perform an estimation of the discount rate over time. For this purpose, the Danish offshore wind farms Anholt [10], Horns Rev 3 [11] and Kriegers Flak [12] are used. In addition, to obtain further calibration points, the nearshore wind farms Vesterhav Syd [14] and Vesterhav Nord [13] are added, by reducing the distance to shore threshold from 10 km to 4 km temporarily. The locations of the mentioned wind farms are marked on figure 1.2. Table 5.1 presents the characteristics of the 5 wind farms (upper part of table until first dashed line), which are fed into the LCoE-model in order to determine the CoE for each wind farm using the methodology and the input parameters described in chapter 3 and 4. The output obtained by the model including wind resources, annual energy production and capital and operational expenses is shown in the sec- ond part of table 5.1 (until second dashed line). The last part of the table presents the computed discount rate which is used to reach the LCoE provided by the DWIA [60]. The wind resource data at 100 m height is used for all wind farms, apart from Anholt where 80 m is applied. The (SG) turbine (SG 8.0-167 DD) [61], which is to be installed at Kriegers Flak and Vesterhav Nord and Syd, does not have a public accessible power curve. Thus, the Vestas V164-8.0MW turbine’s power curve is scaled up to 8.4 WM for the above mentioned wind farms and to 8.3 WM for Horns Rev 3 by extending the power curve to rated power levels of 8.3 and 8.4 MW, respectively. It has to be noted that this is not the most accurate procedure, but as no other data is available for the SG turbine and the data is used to roughly estimate the discount rate, the procedure is assumed to be a good approximation.

The outcome of the analysis reveals the expected decrease in CAPEX per MW installed, from approximately 3 Me/MW for Anholt to 2.4 Me/MW for Horns Rev 3 and 2.2 Me/MW for Kriegers Flak. These numbers correspond relatively well with the values stated in the technical report provided by the DEA [6] (2.46 Me/MW for Horns Rev 3, 1.81-2.13 for Kriegers Flak and 2.07 Me/MW for the nearshore projects Vesterhav Nord and Syd). However, the model seems to underestimate the expenses for Horns Rev 3. This is as well reflected in the resulting discount rate of 11.6%, close to the obtained rate for Anholt. The great drop in prices from Horns Rev 3 to the Kriegers Flak project is not captured by the model, as a smother linear reduction of expenses is assumed from 2008-2018. 5.2 Calibration of discount rate 41

Table 5.1: Key parameters of the Danish wind farms Anholt, Horns Rev 3 (HR3), Kriegers Flak (KF), Vesterhav (VH) Nord and Syd (until first dashed line), includ- ing obtained wind resource data, water depth, distance to PCC (point of common coupling), CAPEX, OPEX and Net AEP (second part, from first dashed line to sec- ond). All calculations are performed with a lifetime of 25 years, a farm layout using turbine spacing of 7×rotor diameter and wind resource data at 100 m height and 80 m (Anholt only). LCoE given by [60] utilized to determine the discount rate (last part).

Anholt HR3 KF VH Nord VH Syd

Year of tender 2010 2015 2016 2016 2016 Year of FID 1 2011 2016 2018 2017 2017 Wind farm developer Ørsted Vattenfall Vattenfall Vattenfall Vattenfall Farm cap. [MW] 399.6 406.7 604.8 168 176.4 Turbine cap. [MW] 3.6 8.3 8.4 8.4 8.4 Rotor diameter [m] 120 164 167 167 167 Water depth [m] 2 16 19 18 22 18 Distance to PCC [km] 68 63 62 38 42 Mean wind speed [ms−1] 3 8.73 9.08 8.71 8.89 8.83 Weibull A [ms−1] 3 10.29 10.63 10.19 10.55 10.61 Weibull k [-] 3 2.24 2.21 2.17 2.27 2.27 CAPEX total [Me] 4 1,207 963 1,342 376 357 CAPEX [Me/MW] 3.02 2.37 2.22 2.13 2.12 OPEX [e/MWh] 5 25.25 16.78 15.09 15.93 15.93 Net AEP [GWh/year] 6 1,561 1,560 2,174 650 612 LCoE [e/MWh] 7 111.0 85.0 56.9 55.1 55.1 Discount rate [%] 11.7 11.6 5.2 5.2 5.2

1 FID: Final investment decision. 2 Water depth averaged over wind farm area. 3 Wind resource data averaged over wind farm area. A and k are Weibull scale and shape parameters, respectively (see section 4.1). 4 CAPEX: Capital expenditures. Obtained from turbine cost in figure 4.5, substructure cost in 4.6, installation cost in section 4.3.3 and grid connection according to section 4.3.4. 5 OPEX: Operational expenditures. Obtained from figure 4.12. 6 Net AEP: Annual energy production after losses, with 13% losses. Determined using method in section 4.1. 7 Given by Danish Wind Industry Association [60]. 42 5 Results

A discount rate above 10% for Anholt seems reasonable since the offshore wind industry was still under development at that time. This is associated with higher risk and thus greater discount rates (IRENA discount rate in 2001 equals 12% [5]). The discount rates for Kriegers Flak and the nearshore projects seem to be underestimated slightly, as the discount rate in Denmark is currently expected to be in a range of 6-10% (according to COWI’s information obtained from the industry [2]). Also a discount rate study for various renewable energy generation units undertaken by Grant Thornton in 2017 [62] finds a discount rate for offshore wind energy of 7-8%in the Nordic countries (Denmark, Norway, Sweden and Finland). To keep the model simple, linear reductions of the discount rate are assumed, starting from 12% in 2008, dropping to 7% in 2018 and further to 6% in 2030. Though, a significant uncertainty and variation from project to project is coupled to this parameter, here indicated by a discount rate ±2% throughout the entire time frame. In 2030 the uncertainty range is expected to lie between 4% and 8%, based on COWI’s report [2] (8%), DEA’s report [6] (4.5%) and IRENA’s prediction of 9% in 2015 and 7.5% in 2030 [5]. The latter discussed discount rate development including uncertainties is shown in figure 5.2, where the baseline scenario is used as input for the LCoE-model.

Figure 5.2: Development of the discount rate from 2008-2030 with uncertainty, in- cluding model derived discount rates (table 5.1) of existing or planned wind farms (Anholt [10], Horns Rev 3 [11], Kriegers Flak [12], Vesterhav Syd [14] and Nord [13]), COWI’s assumed rate of 8% [2], DEA’s of 4.5% [6] and IRENA’s discount rate [5].

5.3 Levelized Cost of Energy (LCoE)

This section presents the resulting LCoE for the entire period under study, 2008- 2030, taking technology development into consideration. It is assumed that the final investment decision (FID) for the three new Danish offshore wind farms in the period 5.3 Levelized Cost of Energy (LCoE) 43 until 2030 takes place 2 years after the tender. As the DEA has set the dates of tender to 2019/2020, 2021 and 2023, the expected years of FID become 2021, 2023 and 2025 for the first, second and third planned wind farm of each 800 MW, respectively. The LCoE-map of the first wind farm to be tendered in 2019/2020 and withthe year of FID in 2021 is illustrated in figure 5.3. The map showcases the LCoE across the entire map (excluding water depth >50 m and distance to shore <10 km), though the boundary of the Danish Sea is indicated with a black dashed line and only this area is used for further analysis.

Figure 5.3: Levelized Cost of Energy (LCoE)-map for the first 800 MW wind farm, tendered in 2019/2020 and assumed year of FID (final investment decision) in 2021. Input parameters: Vestas V164-8.0MW turbine, lifetime of 30 years, AEP (annual energy production) determined using method in section 4.1, CAPEX (capital expen- ditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12 and the discount rate from figure 5.2. Areas closer than 10 km to shore marked light grey, areas with water depth deeper than 50 m marked dark grey, boundary of the Danish Sea indicated through black dashed line and the selected sites by the Danish Energy Agency (DEA) [1] surrounded by a purple line.

The LCoE-map ranges from values of approximately 47 to 81 e/MWh, revealing the significant importance and impact of the discount rate, since the LCoE compared to the CoE in figure 5.1 has nearly doubled. The variables affecting the LCoE are 44 5 Results water depth which is coupled to the substructure cost and the wind resources coupled to the annual energy production (AEP) of the wind farm. In addition, the distance to the closest point of common coupling (PCC) affects the cable cost and thus the LCoE. Generally, a strong dependency of the LCoE in the above presented figure 5.3 and the water depth in figure 4.3 can be observed. The shape of the bathymetry is clearly displayed in the LCoE distribution, where lower LCoE is obtained for shallower water. Moreover, the most profitable areas in the Danish sea are within short distances from the coast, since cable costs are kept on a low level in these areas and the water is generally shallower. The spots around the coast of Jutland result in the lowest LCoE, especially around the Horns Rev area. This is due to a combination of high wind speeds in the North sea (see figure 4.1) and relatively low water depth (figure 4.3). As a consequence of these factors, the area in Jammerbugt is very attractive as well. Also east of northern Jutland (in Kattegat, close to Anholt offshore wind farm) low LCoE is found which is caused by very shallow water and still relatively high wind speeds. However, considerable are as well the areas east of Sealand (around Kriegers Flak) due to shallow water and a relatively close by PCC (Bjæverskov). Though, the LCoE is higher than in the North Sea caused by lower wind speeds. LCoE around Bornholm is among the highest as large distances to the closest grid connection point (East Sealand) and deep water occur here. Additionally, wind speeds are lower in these areas, resulting in lower LCoE also in shallow water (Southwest of Bornholm). Summing up, the LCoE output of the developed model is influenced by all vari- ables such as water depth, wind speed and distance to PCC. However, a notable influence through the water depth is observed since the shape of the bathymetry is clearly visible in the LCoE-distribution across the Danish Sea.

A LCoE value is assigned to a total number of 205,173 pixels within the limits of the Danish Sea (black dashed line in figure 5.3). With an area of 500×500 m per pixel, this corresponds to 51,293.25 km2 of sea area. The remaining pixels are either outside the boundary of the Danish Sea, too close to shore (<10 km from shore), too deep (>50 m water depth) or onshore. In order to analyze the occurrence frequency of the LCoE throughout the Danish Sea, figure 5.4 displays a histogram of the relative occurrence frequency (left) as well as the cumulative occurrence frequency (right) of the considered LCoE values in the latter LCoE-map (figure 5.3). The left figure highlights the quartiles Q1, Q2, Q3 and Q4. The number behind the quartile indicates the percentile which is used to obtain the respective quartile, meaning that Q1(0.25) states that 25% of the area (or LCoE values) is lower than or equal to the respective obtained LCoE. The quartile Q2 indicates the median of all values while Q4 is the maximum value. 5.3 Levelized Cost of Energy (LCoE) 45

Figure 5.4: Relative frequency and cumulative relative frequency of occurrence of LCoE across the Danish Sea (total considered area 51,293.25 km2) in the year 2021 of final investment decision (FID). (b) showing the 4 quartiles Q1, Q2 (median),Q3 and Q4 (maximum), meaning respectively 25%, 50%, 75% and 100% ≤ the respective LCoE.

Figure 5.4 reveals that 50% of the area of the Danish Sea - corresponding to nearly 25,650 km2 - represent LCoE-values between approximately 47 and 58 e/MWh. Even 75% of the values are below 62 e/MWh. The remaining 25% are spread between 62 and 81 e/MWh. Furthermore, it can be concluded that less than 5% of the area contains values below 50 e/MWh. These areas are located close to Horns Rev 3 and around the coast of northern Jutland. The majority is represented by values between 55 and slightly above 60 e/MWh. Only a little share of the area (less than 3%) lies above 68 e/MWh. These areas are mainly located far from shore in deep water. Considering the three future wind farms with a turbine spacing of 7×rotor diameter and the Vestas V164-8.0MW turbine, a total area of 1.32 km2 per turbine is needed, corresponding to 395 km2 for all three wind farms (2400 MW). This area can without difficulties be implemented within the cheapest 25% of the Danish Sea, as the quartile Q1 (0.25) equals roughly 12,800 km2. However, it has to be kept in mind that these areas are partly not available for new wind farms due to natural restrictions, land use, other wind farms, shipping routes and many other factors (for further information see Laura Schröder’s MSc thesis from 2016 [63]).

In order to analyze the development of the LCoE throughout the entire time span from 2008-2030, several technological improvements are implemented, considering varying wind turbine and farm capacity as well as life time improvements. Initially, from 2008 and on-wards, the Anholt wind farm characteristics are fed into the LCoE- 46 5 Results model. In 2016, when the FID for Horns Rev 3 was fixed, the turbine capacity increased to 8 MW. As the planned three wind farms are going to be designed with a capacity of 800 MW and 30 year lifetime, these changes are applied in 2020. Fur- thermore, a turbine size of 10 MW in 2023 (second new farm) and 12 MW in 2025 (third farm) is utilized. These technological changes are summarized in table 5.2:

Table 5.2: Technological development from 2008-2030. Changed parameter indicated in bold font.

Year 2008 2016 2020 2023 2025

Farm capacity PF [MW] 400 400 800 800 800 Turbine capacity PT [MW] 3.6 8 8 10 12 Lifetime LT [years] 25 25 30 30 30

The LCoE-maps considering the above mentioned varying input parameters are resulting in the same color distribution across the entire map in every year. Exclu- sively the range of LCoE, meaning the color bar scale is changing. Therefore, only the range of the LCoE for each year is illustrated in figure 5.5, including the above mentioned technological developments. In addition, the quartiles of the LCoE occur- rence over the Danish Sea are shown for each year, the median (Q2) is indicated by a red line. 5.3 Levelized Cost of Energy (LCoE) 47

Figure 5.5: LCoE range across the Danish Sea from 2008-2030, considering different wind farm design developments, with the variables wind farm capacity PF , wind turbine capacity PT and lifetime LT . The bars for each year show the 4 quartiles Q1, Q2 (median), Q3 and Q4 (maximum), meaning respectively 25%, 50%, 75% and 100% ≤ the respective LCoE. The upper and lower boundaries of LCoE indicated by black line.

Figure 5.5 clearly displays the drop in LCoE over the entire period 2008-2030, especially the steep slope from 2008-2015, consistent with the in chapter 4 indicated reductions in CAPEX, OPEX and the discount rate. In 2016 this steep slope flattens partially out, due to a flattening OPEX-curve (figure 4.12). The same happens in 2018 when the CAPEX-curves (figure 4.5 and 4.6) start to decrease more slowly. The changes in turbine and farm capacity as well as lifetime clearly induce LCoE reductions, especially when introducing a lifetime of 30 years and a farm capacity of 800 MW in 2020. Also, increased turbine capacities in 2023 and 2025 cause significant drops in LCoE. Furthermore, the spread of the LCoE in Danish waters is decreasing, due to past and future price reductions - moving from a range of approximately 115-200 e/MWh in 2008 to 36-60 e/MWh in 2030. The same distribution of LCoE occurrence in the Danish Sea as analyzed in figure 5.4, occurs in each year in figure 5.5. 48 5 Results

5.4 Uncertainty and sensitivity

The estimation and prediction of the LCoE for offshore wind power are subject to great uncertainty coming from several sources. On the one hand, uncertainty is associated with the input data such as wind resources data, CAPEX, OPEX and the discount rate. On the other hand, the methodology used in the LCoE-model, considering all assumptions and simplifications, implies uncertainty on the LCoE as well. This includes, among others, the way of computing the LCoE (equation 3.1) by assuming the OPEX to be a fixed cost per MWh as well as the process of averaging the water depth and AEP over the wind farm area when computing the LCoE for one pixel. Also, the loss estimation and the assumption regarding turbine and substructure installation time and cost are subject to uncertainty. Furthermore, national and international political decisions and regulations as well as market fluctuations can have significant influence on the LCoE. It is assumed that the economical aspects such as CAPEX, OPEX and especially the discount rate have a considerable influence on future LCoE-predictions. Therefore, the LCoE range across the Danish Sea is computed considering a high cost scenario and a low cost scenario for 2030, as the uncertainty is highest at the end of the predicted period. For that purpose the CAPEX, the OPEX and the discount rate are set to a minimum and maximum level. The uncertainties described in chapter 4 are used:

• CAPEX: ± 10% • OPEX: ± 10% • Discount rate: ± 33% (discount rate value: low 4%, baseline 6%, high 8%)

The LCoE ranges for the baseline, high cost and low cost scenario are displayed in the following figure 5.6. The figure includes the relative occurrence frequency ofthe LCoE across the Danish Sea and the corresponding quartile analysis (according to the procedure described in the previous section 3.1). 5.4 Uncertainty and sensitivity 49

Figure 5.6: LCoE uncertainty for baseline, high cost and low cost scenario in 2030, considering CAPEX ± 10%, OPEX ± 10% and discount rate ± 33%. Relative occur- rence frequency of LCoE across the Danish Sea (left) and quartile analysis (right).

Figure 5.6 illustrates that the distribution of LCoE occurrence is the same for the 3 scenarios. However, the range of LCoE distribution becomes greater for higher costs due to the relative higher share of expenses in more expensive areas. Generally, the right graph in figure 5.6 displays clearly the great influence of CAPEX, OPEX and discount rate uncertainty on the LCoE. Considering Q2 (median/ 50% most profitable areas of Danish waters), the difference in LCoE from the baseline scenario (45 e/MWh) to the high and low cost scenario becomes approximately +12 e/MWh and -10 e/MWh, respectively. The relative uncertainty for each quartile amounts to around +27% in the high cost and -23% in the low cost scenario, compared to the baseline scenario. This is a significant range of uncertainty, though it has to be kept in mind that these are the results for a case where all economical input parameters are driven to the minimum or maximum of the given uncertainty range at the same time. This might not be the case in real-life. Nevertheless, cost and risk associated with offshore wind energy projects are currently under fast development. This makes the prediction of the economical parameters highly uncertain, as it is not known how much the cost can be driven down by the industry and/or government. In order to analyze the effect of various parameters such as cost, discount rate and lifetime on the LCoE, a sensitivity analysis is presented in the following. For that purpose a wind farm in the North Sea (pixel of center of the wind farm: (UTM easting, northing)=(418500 m, 6246500 m)) is designed, with an average water depth of 28 m, a distance to the closest PCC of 50 km and a lifetime of 30 years. The parameters used are summarized in table 5.3: 50 5 Results

Table 5.3: Key parameters of the wind farm considered for the sensitivity analy- sis in final investment decision (FID) year 2021. AEP (annual energy production) determined using method in section 4.1. PCC (point of common coupling to grid).

Parameter Value Center pixel (UTM easting, northing) [m] (418500, 6246500) Average water depth [m] 28 Distance to PCC [km] 50 Turbine type V164-8.0MW Turbine capacity [MW] 8 Farm capacity [MW] 800 Gross AEP [GWh/year] 3,620 Lifetime [years] 30 Losses [%] 13

The contribution of each wind farm element to the entire LCoE for the above described wind farm is shown in the next figure:

Figure 5.7: Contribution of each wind farm element to LCoE in the year of final investment decision (FID) in 2021 (wind farm used for sensitivity analysis). Input parameters can be found in table 5.3, cost parameters in chapter 4: CAPEX (capital expenditures) from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12, AEP (annual energy production) determined using method in section 4.1.

Comparing the results of the above presented pie chart with the analysis performed by IRENA in 2017 [5] (considering European price levels), the contribution of turbine, 5.4 Uncertainty and sensitivity 51 substructure, and electrical connection cost (Array cable plus Transmission) is very similar. Though, the contribution of OPEX in this study is estimated higher than in the IRENA report (19%) while the installation cost is significantly lower in this study (IRENA: 20%). Also comparing to other studies such as NREL’s report [64], the installation cost in this study seems to be underestimated, which might be due to the fact that this study does not take site conditions, distance to the closest port nor labour cost into account. For the sensitivity analysis, different cost parameters have been varied with ± 10% of the original cost considered in the baseline scenario. Among these are the turbine cost, substructure cost, cost for planning and development, the grid connection cost (inter-array cable plus transmission cost consisting of offshore and onshore substation and transmission cable) and OPEX. Additionally, the discount rate is varied by ± 10%, while a lifetime of 25 to 35 years is taken into account. The resulting relative changes in LCoE are shown in figure 5.8:

Figure 5.8: LCoE sensitivity to cost parameters, discount rate and lifetime with the specifications stated in table 5.3 and final investment decision in 2021. Cost parameters can be found in chapter 4: CAPEX (capital expenditures) derived from figure 4.5, 4.6 and section 4.3.4. OPEX (operational expenditures) obtained from figure 4.12 and the discount rate from figure 5.2.

As already discussed in section 5.2, the discount rate is a major factor contributing to the LCoE as figure 5.8 displays. The change of ± 10%, corresponding to a discount rate of approximately 6.2 to 7.5% (baseline 6.8%), causes a change in LCoE of ±5%. 52 5 Results

The lifetime turns out to contribute considerably to changes in LCoE as well, having differing negative or positive influence on the LCoE due to the appearance ofthe lifetime in the levelizing factor computation (equation 3.4). Taking a look at the sensitivity of the LCoE to the CAPEX (turbine, substructure, planning and grid connection cost) in figure 5.8, the turbine cost appears to have a great impact on the LCoE, which corresponds to the fact that the turbine takes the largest share of the LCoE as shown in figure 5.7. This is followed by the sensitivity of LCoE to the substructure cost. Planning and development expenses as well as the grid connection cost have a rather minor influence on the LCoE compared to the other parameters. On the contrary, the OPEX has a significant impact on the LCoE, as it represents a considerable share of the LCoE. Thus, when aiming to reduce the LCoE of offshore wind power, the main factors driving down the cost of energy are reduced discount rate and increased lifetime. Also, decreasing the turbine cost and the OPEX translate into significant LCoE reductions. As concluded in the previous section 3.1, also technology improvements such as larger turbines that are increasing the AEP, help to decrease the LCoE.

The last part of this section aims to investigate the uncertainty arising from the process of averaging the water depth and wind resources over the entire area of the wind farm under study (for a specific pixel on the LCoE-map). For that purpose the wind farm described in table 5.3 is used. The water depth varies with ± 9% over the wind farm area, causing a variation in the LCoE of ±1.6% whereas the AEP across the wind farm area only varies with ± 0.8% resulting in an LCoE ± 0.6%. Both, water depth and wind resources translate into uncertainties of less than ±1 e/MWh, thus significantly lower than uncertainties related to the economical aspects. Noteworthy is as well that the averaging process is done consecutively across the entire map for each pixel, thus taking the regional variations of water depth and wind conditions into account.

5.5 Selected sites

After a large-scale screening process, the Danish Energy Agency (DEA) published suitable sites for future offshore wind farms in September 2018[1], resulting in four areas - one in the North Sea, one in the Jammerbugt, one close to Hesselø and one around Kriegers Flak (all marked in figure 1.2). No screening process is conducted in this study, since a screening was already preformed by the DEA and Laura Schröder in her MSc thesis [63] in 2016. Additionally, the DEA hired COWI A/S in Septem- ber 2018 to perform a detailed screening of the selected sites. The screening process depends on many restricting parameters related to geographical and geological condi- tions, land use interests and nature. This includes, among others, parameters such as seabed conditions, water depth, shipping routes, air traffic and radars, military and mines, oil and gas activities, submarine cables, fishing as well as nature protected areas and animals [63]. The screening process including the environmental impact 5.5 Selected sites 53 assessment (EIA) is a very important aspect in the planning and development process of a wind farm, though not part of this study. COWI’s report regarding the detailed screening and performance analysis of the four selected sites was published mid December 2018. Therefore, it is decided to conduct the planning of one offshore wind farm at each site and rank them according to their LCoE. These results are finally compared with COWI’s findings in [2].

The following LCoE comparison is performed for the first planned offshore wind farm tendered in 2019/2020 and assumed final investment decision (FID) in 2021. The software WAsP [36] is used to conduct the wind resource assessment for each pre-selected site, with wind resource data taken from the Global Wind Atlas [34]. Wake analysis are carried out in the software Fuga [57]. The net AEP is obtained after considering wake losses and the losses given in table 4.6 (3% availability losses and 1.5% electrical losses). A farm capacity of 800 MW and the Vestas 8 MW turbine (V164-8.0MW [33]) are used for the computations. Usually, the layouts of wind farms are optimized in order to reduce wake losses. However, as the purpose of this analysis is to compare the LCoE of different sites, no optimization is performed. Instead, layouts with turbine spacing of 7×rotor diameter and a rectangular arrangement of turbines (if possible) are designed. A minimum distance to shore of 20 km, as required by the DEA, is fulfilled for all wind farms. The economic input data (cost ofwind farm components, OPEX and discount rate) presented in chapter 4 is used for the year of FID 2021 as well as a lifetime of 30 years.

Nordsøen

Nordsøen is divided into two zones - a large area Nordsøen A and a smaller area further from shore Nordsøen B. Comparing the wind resources in these two areas, the difference is not considerable compared to the larger distance to shore that comes along with higher transmission cable cost. The mean wind speed at the most western part of Nordsøen A does not differ from the mean wind speed in Nordsøen B. There- fore, it is decided not to include Nordsøen B in the analysis. Also the southern part of Nordsøen A is not suitable for offshore wind farms since oil and gas pipelines as well as the electrical connection between Denmark and the United Kingdom, Viking Link, are planned to run through this area (COWI [2]). Thus, this area is neglected as well. The remaining area, suitable for offshore wind farms, is from now on named ’Nordsøen’. The most optimal position for a single wind farm in Nordsøen occurs in the north- ern part, as it is close to the next grid connection point and close to shore. The water depth is of minor importance as it does not differ notable throughout the entire area of Nordsøen (see figure 4.3). The wind resources for the northern part of Nordsøen are shown in figure 5.9 (obtained in WAsP): 54 5 Results

Figure 5.9: Nordsøen wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived).

The wind rose in figure 5.9 reveals that the prevailing wind directions occur from west-northwest to south-southwest, moving counterclockwise, with a mean wind speed of 9 ms−1. In order to minimize wake effects, the majority of the wind turbines face the free incoming wind from west, still keeping in mind a tight layout (7×rotor diameter) to reduce inter-array cable expenses. The chosen layout of 5×20 turbines and the position within the area Nordsøen is shown in figure 5.10:

Figure 5.10: Nordsøen wind farm layout (dots) and disregarded zone (black dashed area) due to planned oil and gas pipelines and electrical interconnection between Denmark and the UK (Viking Link). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km.

The determined results for the wind farm in Nordsøen are presented in the follow- ing table 5.4: 5.5 Selected sites 55

Table 5.4: Nordsøen offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Idomlund). The inter-array cable length obtained by connecting 7 turbines (accord- ing to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area.

Inter-array Transmission LCoE LCoE-map Mean water Wake loss Net AEP cable length cable length [e/MWh] [e/MWh] depth [m] [%] [GWh/year] [km] [km] 53 54 27 8.34 3,166 180 43

The LCoE obtained for the planned wind farm (figure 5.10) results in 53 e/MWh, while the LCoE-map in figure 5.3 displays approximately 54 e/MWh. This difference is caused by higher wake losses applied in the LCoE-model (9%). Nonetheless, the results are very similar. The wake losses within the planned wind farm can be reduced by performing a layout optimization, making this area even more profitable.

Figure 5.10 clearly displays the huge space available for offshore wind farms in Nordsøen. It is convenient to erect several wind farms inside the area, as wind speeds are high. Additionally, grid connection cost is reduced as soon as one wind farm is connected. The area suitable for construction of wind farms has an area of approximately 1500 km2. Considering the 8 MW Vestas turbine with a rotor diameter of 164 m and a spacing between turbines of 7×rotor diameter, the area could cover a total installed capacity of approximately 9 GW. However, this does not consider the need for spacing between wind farms nor the maximum capacity that can be connected to a 400 kV grid connection point. Nevertheless, Nordsøen has a great potential for future offshore wind farms.

Jammerbugt

The area Jammerbugt has a prevailing wind direction coming from west to south- southwest and a relatively high mean wind speed of 8.8 ms−1, as the wind rose and Weibull probability density function in the next figure 5.11 demonstrate: 56 5 Results

Figure 5.11: Jammerbugt wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived).

Though, the area had to be extended by COWI and the DEA, as it was placed too close to shore - the closest boundaries of the area were placed 15 km from shore instead of the required 20 km. The updated area [2] is illustrated in figure 5.12 together with the proposed layout to fulfill the 20 km distance to shore. The layout is not placed optimally considering the prevailing wind direction and length of inter- array cables. However, to satisfy the 20 km distance to shore, there is not much room for a layout optimization.

Figure 5.12: Jammerbugt wind farm layout (dots), original area (orange line) pub- lished by the DEA [1] and updated area (grey polygon) [2]. Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km.

The resulting LCoE including the parameters used for the computation is shown in table 5.5: 5.5 Selected sites 57

Table 5.5: Jammerbugt offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Ferslev). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area.

Inter-array Transmission LCoE LCoE-map Mean water Wake loss Net AEP cable length cable length [e/MWh] [e/MWh] depth [m] [%] [GWh/year] [km] [km] 53 53 20 9.31 3,054 190 54

A resulting LCoE of 53 e/MWh corresponds very well with the outcome of the LCoE-map (figure 5.3). The LCoE-result is very similar to the one obtained for Nordsøen. Even though the wake losses are significantly higher for Jammerbugt, the shallower water is compensating for the lower Net AEP and the longer distance to the onshore grid connection point.

Hesselø

The Hesselø area is consisting of 2 zones - Hesselø A and Hesselø B. It is decided not to include Hesselø A (the small area in figure 1.2) as it only comprises 43 km2 - barely enough for 260 MW installed capacity, considering the Vestas V164-8.0MW turbine and a spacing of 7×rotor diameter. A wind farm is designed in the southern part of the area Hesselø B (from now on called Hesselø) since the water is shallower here and the distance to the next grid connection point (Trige) is closer by than in the northern part of the area. A squared layout with a spacing of 7×rotor diameter is used (figure 5.14). The dominant wind direction is west-northwest to south-southwest moving counterclockwise and the mean wind speed amounts to approximately 8.7 ms−1, as figure 5.13 illustrates.

Figure 5.13: Hesselø wind rose (left) and Weibull probability density function (right) at 100 m height (WAsP derived). 58 5 Results

Figure 5.14: Hesselø wind farm layout (dots). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km.

The obtained LCoE including all calculation parameters is displayed in table 5.6. The LCoE of the designed wind farm is slightly exceeding the LCoE computed by the LCoE-model (figure 5.3), due to higher wake losses. Performing a layout optimization is definitely recommended here, as there is sufficient space to even construct twowind farms in the area Hesselø.

Table 5.6: Hesselø offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of common coupling (Trige). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area.

Inter-array Transmission LCoE LCoE-map Mean water Wake loss Net AEP cable length cable length [e/MWh] [e/MWh] depth [m] [%] [GWh/year] [km] [km] 58 57 26 9.47 2,981 180 88

Kriegers Flak area The area Kriegers Flak is as well divided into two zones, Kriegers Flak A and B. In this case though, it is necessary to use both zones to obtain an installed capacity of 800 MW, as the areas only consist of 66 and 78 km2, respectively for zone A and B. The wind speeds are slightly higher for Kriegers Flak B (v¯ = 8.74 ms−1) than for A 5.5 Selected sites 59

(v¯ = 8.62 ms−1). However, the wind roses are very similar. Therefore, only the wind resources for Kriegers Flak A are shown in this section (figure 5.15), for Kriegers Flak B see Appendix A.1. The wind blows most frequently from west and west-southwest, though also occasionally from east.

Figure 5.15: Kriegers Flak A wind rose (left) and Weibull probability density function (right) at 100 m height. (For Kriegers Flak B see Appendix A.1)

The layouts are designed with 53 turbines in zone A and 47 turbines in zone B, equidistantly spaced with 7×rotor diameter, as shown in the following figure. Again, no design optimization is performed. However, it might be more profitable to fill out the whole area with turbines by applying a larger spacing between the turbines.

(a) Kriegers Flak A. (b) Kriegers Flak B.

Figure 5.16: Kriegers Flak wind farm layout (dots) for zone A (53 turbines) and zone B (47 turbines). Turbine spacing 7×rotor diameter = 7 × 164 m = 1.15 km.

Table 5.7 presents the obtained modelling results in WAsp/Fuga as well as the used input parameters and finally the computed LCoE for both areas. 60 5 Results

Table 5.7: Kriegers Flak A and B offshore wind farm key parameters and results. Transmission cable length is the distance from substation to closest point of com- mon coupling (Bjæverskov). The inter-array cable length obtained by connecting 7 turbines (according to section 4.3.4) to each cable, then leading to the substation. The obtained Levelized Cost of Energy (LCoE) indicated with bold font. LCoE-map determined from figure 5.3, averaged over wind farm area.

Inter-array Transmission LCoE LCoE-map Mean water Wake loss Net AEP cable length cable length [e/MWh] [e/MWh] depth [m] [%] [GWh/year] [km] [km] KF A 53 56 28 6.01 1,665 85 40 KF B 56 56 27 7.84 1,441 70 67

The LCoE-result for Kriegers Flak B corresponds very well with the LCoE shown on the map in figure 5.3. However, the one obtained for Kriegers Flak A is significantly lower than the LCoE-model’s computation, which is caused by notable lower wake losses than assumed in the model. Combining the expenses and the annual energy production of both farms, a LCoE of 54 e/MWh is found. It has to be noticed that this result might be underestimated, as two offshore substations are needed for the Kriegers Flak A+B project. The model does not account for the most likely occurring higher costs related to the installation process of two substations instead of one.

Ranking of sites Finally, the above analyzed offshore wind farms are ranked according to their obtained LCoE and compared to COWI’s ranking [2]. For better comparison the LCoE in table 5.8 is as well presented with a discount rate of 8%, corresponding to the one used by COWI. Table 5.8: Ranking of selected sites according to LCoE in final investment decision year 2021 and compared with COWI’s ranking in [2].

LCoE [e/MWh] LCoE [e/MWh] COWI [e/Mwh] No. Site (r=6.76%) (r=8%) (r=8%) 1 Jammerbugt 53 58 64 2 Nordsøen 53 58 60 3 Kriegers Flak 54 59 64 4 Hesselø 58 63 62

The ranking obtained in this study places Jammerbugt on the first place, very closely followed by Nordsøen. Both areas are well-placed in the North Sea where wind speeds are high and stable. Also grid connection points are close by. On the third place follows Kriegers Flak not far behind Jammerbugt and Nordsøen. Low wake losses are compensating for lower wind speeds in the Kriegers Flak area com- pared to wind speeds in the North Sea. However, as discussed above, the LCoE for 5.5 Selected sites 61 the Kriegers Flak area might be too low due to an underestimation of the substation costs. The most expensive area, according to this analysis, is Hesselø. It is appearing relatively far behind the other sites, due to larger wake losses and a long distance to the closest PCC.

Comparing the obtained results with COWI’s LCoE [2], the overall order of mag- nitude of LCoE is relatively similar, though COWI’s numbers are slightly higher. However, COWI’s ranking differs from the one presented in this study. This dis- similarity arises mainly through COWI’s detailed analysis of the grid connection cost since the Danish transmission system operator Energinet provided them with internal information. This study, on the other hand, models the grid connection cost simi- lar for each site without considering the conditions of the onshore connection points. COWI clearly places Nordsøen as the most profitable site, followed by Hesselø. The low LCoE for Hesselø in COWI’s analysis compared to the high LCoE obtained in this study, results from different grid connection points. COWI suggests to couple the wind farm to the grid at the northern part of Sealand, whereas the wind farm in this report is connected to Jutland (Trige). Lastly Jammerbugt and Kriegers Flak are ranked with the highest LCoE. Kriegers Flak is rather expensive due to the fact that two substations are needed whereas Jammerbugt requires a costly expansion of the onshore grid.

It can be concluded that the developed model and the latter performed study of the pre-selected sites result in relatively similar LCoE values compared to COWI’s study. Thus, considering the input parameters of this study, reasonable LCoE-results are obtained. However, the ranking differs due to COWI’s detailed inclusion of grid connection cost, that affects the ranking of sites according to their LCoE notably. Overall, the obtained LCoE values for the different sites differ not much. Hence, the matter of driving down the cost of offshore wind energy is not necessarily related to the selection of the site - as long as sites relatively close to shore are still available - but rather has to come from technology improvements and reductions of economical parameters or from improvements in the supply chain and installation campaigns. 62 5 Results

5.6 Electricity price comparison

The following section provides an analysis to assess the feasibility of subsidy-free offshore wind projects in Denmark in the future. For that purpose, the electricity price prediction undertaken by the Danish Energy Agency (DEA) in 2017 [3] is compared to the LCoE-results presented in figure 5.5. It is investigated if the electricity price would exceed the LCoE of offshore wind energy in the future so that wind farm operators are able to make profits on the electricity spot market. The DEA states three main factors affecting the electricity price in the future:

• Interconnection between Denmark and neighbouring countries.

• Fossil fuel prices.

• Carbon credits1 prices.

Generally, according to DEA’s baseline scenario, the Danish electricity price is ex- pected to increase in the future. Since Denmark’s electricity price is primarily de- termined on Nord Pool’s 2 day-ahead market and Denmark mainly is a price-taker, the Danish electricity price is strongly influenced by neighbouring country’s genera- tion. It is assumed that the generally lower electricity price in Denmark will converge towards neighbouring country’s higher prices due to the planned expansion and im- provements of interconnections between Denmark and other countries. Planned power connections are the subsea cable COBRA, linking Denmark and the Netherlands, ex- pected to operate from 2020 on, as well as the link between Denmark and Germany through the offshore wind farm Kriegers Flak, operating in 2019. Additionally, the Viking Link is supposed to connect Denmark to the United Kingdom from 2022 on. As electricity consumption is assumed to increase from 2020 on-wards, the DEA estimates likewise a rise in fossil fuel consumption and thus increasing electricity prices. This is as well caused by a reduction of wind energy in the energy system portfolio, as a large share of land-based wind turbines will reach their end of life. However, noteworthy is here that the analysis from 2017 [3] was performed before the new energy agreement was signed in 2018, thus not considering the establishment of the new offshore wind farms. Furthermore, an increase in carbon credit prices is assumed until 2030, as the surplus of certificates created during the financial crises is expected to decrease inthe future due to regulations introduced by the European Commission. The baseline electricity price forecast given by the DEA as well as their predicted high and low cost scenarios are shown in the following figure 5.17. The electricity price is compared to the results obtained by the LCoE-model presented in figure 5.5. Additionally, the year of tender, expected FID and expected year of commissioning of the three new offshore wind farms is indicated. The year of FID is set2years

1 Carbon credit: permit or certificate to emit a certain amount ofCO2 or other greenhouse gases (mass equal to 1 ton of CO2), tradable under European Emissions Trading System (EU ETS) [65]. 2Nordic electricity spot market operator. 5.6 Electricity price comparison 63 after the tender while the year of commissioning is expected to follow 5 years after the tender.

Figure 5.17: Electricity price prediction including uncertainty reproduced from [3] compared to the LCoE for offshore wind across the Danish Sea from 2018-2030 pre- sented in figure 5.5. The years of tender (triangle), expected final investment decision (FID) (circle) and expected year of commissioning (5 years after tender) (square) for the three planned offshore wind farms (1st farm ’red’, 2nd farm ’orange’, 3rd farm ’green’) shown in lower part of the figure.

Figure 5.17 clearly demonstrates that none of the three planned offshore wind farms will be able to operate without subsidies if the electricity price will follow the lowest electricity price scenario while LCoE for offshore wind will develop according to the LCoE-model’s baseline scenario. Not even the low LCoE scenario will reach the low electricity price. However, if electricity prices are rising according to DEA’s baseline scenario, wind farms installed in the cheapest areas could operate without subsidies from 2025 on-wards. Half of the entire area of the Danish Sea (median) could operate subsidy-free from 2028 on-wards. Thus, wind farm 2 could operate without support if placed in a very profitable area and wind farm 3 if placed in the 50% of the sea with lower LCoE - hereby only considering the year of expected commissioning. This would however presume that the wind farm operator already during FID or even during the tendering process knows that electricity prices will rise or LCoE of offshore wind power will decrease. Looking at the high price scenario for electricity, even wind farm number 1 could operate support-free in its commissioning year in 2024 and from 2028 on-wards all wind farms placed in the Danish Sea could run without subsidies. As discussed in section 5.4, the LCoE-results are subject to great uncertainty considering high or low cost scenarios (indicated for 2030 in the 64 5 Results above figure). This implies that the indicated years in which electricity price exceeds the LCoE might vary as well. Nevertheless, the latter discussion only investigates conditions at the date of com- missioning. Looking at the date of the tender - where wind farm operators currently give their bid for the expected need of support - only the third wind farm could exist without subsidies, if erected in the profitable areas and under high electricity price circumstances. Overall, a subsidy-free future for Danish offshore wind power projects is possible if electricity prices will increase notable during the coming years due to rising fossil fuel and/or carbon credit prices or due to a convergence towards neighbouring countries’ higher electricity prices. On the other hand the cost of offshore wind energy would have to experience a reduction - considerably larger than expected in this study - in order to be able to operate subsidy-free under the low electricity price scenario. CHAPTER 6 Conclusion

In this study, a tool was developed that enables an evaluation of the economic attrac- tiveness of future offshore wind energy projects in Denmark until 2030. A methodol- ogy was introduced that provides the Levelized Cost of Energy (LCoE) for offshore wind energy projects across the Danish Sea, based on available data and assumptions. The methodology can beneficially be used by governments and wind farm developers who can adapt the input parameters according to their knowledge and needs. The tool provides a color map illustrating the distribution of LCoE for offshore wind across the Danish Sea considering water depth, distance to shore and wind re- sources. A prediction of future development of economical and technical parameters up until 2030 was performed, hereby including capital and operational expenditures (CAPEX and OPEX), discount rates as well as wind farm capacity, turbine capacity and lifetime of the project.

The analysis of LCoE for offshore wind across the Danish Sea revealed that the most profitable areas are located close to Denmark’s coast. Especially attractive appeared the area around Jutland in the North Sea (west coast of Jutland) but as well areas in the Kattegat northeast of Jutland, yielding in LCoE values of 47-55 e/MWh for the year of final investment decision (FID) in 2021. The low LCoEin these areas is a result of high wind speeds in combination with relatively shallow water. Considerable is as well the area in the Baltic Sea around Kriegers Flak (east of Sealand). Here LCoE values of 52-58 e/MWh were found in 2021 (FID). Overall, a significant drop of LCoE across Danish waters from 2008-2030 was found, moving from a range of approximately 115-200 e/MWh in 2008 to 36-60 e/MWh in 2030. This reduction in LCoE is a result of the assumed decline in CAPEX, OPEX and discount rate. However, a significant drop in LCoE induced by technology improvements was found as well, especially resulting from increased turbine capacity and lifetime improvement. Extending the lifetime from 25 to 30 years in 2020 (FID) while at the same time increasing the wind farm capacity from 400 MW to 800 MW, resulted in a reduction of the LCoE range across the entire Danish Sea from 51-85 e/MWh in 2019 towards a range of 48-82 e/MWh in 2020 (including expected reductions in CAPEX, OPEX and discount rate). Likewise, introducing a 12 MW wind turbine in FID 2025 (10 MW turbine prior) caused a decline in LCoE from 42-73 e/MWh in 2024 to 39-67 e/MWh in 2025. Also, a sensitivity analysis demonstrated that the LCoE is especially sensitive to 66 6 Conclusion the discount rate. Changing the discount rate by ±10% affected the LCoE by ±5%. Also wind turbine cost and OPEX have considerable impact on the LCoE of offshore wind energy projects. Variations of ±10% caused a change in LCoE of ±3% and ±2.5% for turbine cost and OPEX, respectively. Furthermore, it was found that the LCoE predictions up until 2030 are subject to significant uncertainty which is, on the one hand linked to the methodology usedto compute the LCoE. On the other hand, a significant uncertainty is associated with predictions of the economical parameters (CAPEX, OPEX and discount rate). Ac- cording to the baseline scenario of this study, 50% of the entire Danish Sea has a LCoE for offshore wind projects of 45 e/MWh or less in 2030. The associated uncer- tainty was found to be approximately ±11 e/MWh when considering a high and a low cost scenario.

A ranking according to the LCoE of wind farms positioned in the pre-selected areas (published by the Danish Energy Agency [1]) was performed for the year of FID in 2021. The results have shown that the most economically attractive areas are located in the North Sea (Jammerbugt and Nordsøen) with 53 e/MWh, followed by the Kriegers Flak area with 54 e/MWh and lastly Hesselø resulting in 58 e/MWh. Overall, the LCoE values obtained for the pre-selected sites are not significantly dif- ferent. Hence, a reduction in LCoE is not necessarily related to the selection of the site (as long as areas close to the coast are still available), but rather has to result from a reduction in costs as well as improvements of technology, supply chain and installation campaigns.

Ultimately, it was discussed if future Danish offshore wind farms will be able to operate without external subsidies. Therefore, the obtained LCoE range up until 2030 was compared to electricity price predictions provided by the Danish Energy Agency [3]. It was concluded that if prices will follow the low electricity price scenario (around 25 e/MWh), a subsidy-free future for Danish offshore wind energy projects is believed very hard to obtain - not even when considering the low LCoE scenario of this study. However, it might be possible if electricity prices will increase significantly during the coming years. It might be possible for the most profitable areas in the Danish Sea to operate without subsidy from 2025 if electricity prices increase to around 40 e/MWh. This could be achieved through rising fossil fuel or carbon credit prices or by a convergence of Danish electricity prices towards neighbouring countries’ higher electricity prices. CHAPTER 7 Outlook

The present study has shown, that the computed LCoE for Danish offshore wind farms in the future is subject to significant uncertainty. In order to minimize the uncertainties related to the methodology used and the estimation of expenses, one can consider several approaches to improve the quality and accuracy of input data and thus results.

First of all, the used cost estimates in this study are subject to uncertainty since the transparency within the wind turbine industry is relatively poor. However, if pos- sible, the implementation of more accurate costs data would minimize the uncertainty of the model output notable. The expenses related to installation campaigns of substructure and turbines are usually dependent on the weather conditions at site and distance to port. Thus, in order to obtain more accurate estimates for the installation costs, weather conditions such as significant wave height and wind speed as well as suitable harbour locations could be additional model input parameters. The same could be done for the O&M expenses, as these depend on the same factors. Furthermore, the monopile cost function in this study was developed for a 3.6 MW turbine. Though, the monopile cost function should be updated when using larger wind turbines. Overall, the cost for the entire monopile is increasing when using larger turbines, as more material is used for the monopile. However, the cost per MW is expected to decrease. This is partly reflected in the predicted cost reduction for monopiles, but could be done more accurately in order to obtain results that are more realistic. Also different types of foundations such as Jackets or even floating foundations could be included to take deeper water into account as well. Next, analyses have shown that the estimated wake losses and the spacing be- tween the turbines is not optimal. For the purpose of comparison between sites this assumption is reasonable. However, if more precise LCoE results are desired, a design optimization and more detailed wake loss analysis could be included in the model. Another option for improvement could be a more detailed implementation of the transmission system costs by including costs related to onshore cables. Also more detailed cost estimates for each individual onshore plant could be included.

Summarizing, the more precise the input data is, the better the LCoE estimates become. The developed methodology would be an ideal tool for offshore wind farm 68 7 Outlook developers, as they are the stakeholders that have the most information regarding input data, especially the economical aspects.

Ultimately, the application of this tool for other regions could be very interesting, particularly in areas where offshore wind energy is not as deployed as in Denmark. In that way the tool could be used to compare LCoE of offshore wind energy projects across country borders. The model could as well be extended to cover onshore areas by adjusting input parameters. An idea could even be the implementation of the presented methodology in the Global Wind Atlas [34] to show the LCoE across the globe. This would however require local data from other countries e.g. economical aspects or grid connection points. APPENDIX A Additional graphs

A.1 Site selection

Figure A.1 displays the wind rose and Weibull probability density function for the pre-selected site [1] Kriegers Flak B.

Figure A.1: Kriegers Flak B wind rose (left) and Weibull probability density function (right) at 100 m height. 70 A Additional graphs

A.2 Transmission

Figure A.2 presents the selected 400 kV connection points (orange circles) in the Danish grid.

Figure A.2: Danish electricity grid, with marked strong points of common coupling of 400 kV (yellow circles). References

[1] C. Lietzen, Ministry of Energy, Utilities and Climate: Her skal Danmarks næste havvindmøllepark ligge, 2018. [Online]. Available: https://efkm.dk/aktuelt/ nyheder / 2018 / sep / her - skal - danmarks - naeste - havvindmoellepark - ligge/ (visited on Sep. 29, 2018). [2] COWI, “Finscreening af havarealer til etablering af nye havmølleparker - Hov- edrapport,” Tech. Rep. December, 2018. [3] Energistyrelsen, “Fremskrivning af elprisen,” Tech. Rep., 2017. [4] I. Pineda, “Wind Europe: Offshore Wind in Europe, Key trends and statistics 2017,” 2018. [5] IRENA, International Renewable Energy Agency, Innovation Outlook: Offshore Wind. 2017. [Online]. Available: https://www.irena.org/DocumentDownloads/ Publications/IRENA{\_}Innovation{\_}Outlook{\_}Offshore{\_}Wind{\_ }2016.pdf. [6] Danish Energy Agency, “Technology Data for Energy Plants Generation of Electricity and District Heating , Energy,” Tech. Rep., 2016. [7] C. Holbech, “The development of offshore wind - The case of Denmark,” Tech. Rep. February, 2017. [Online]. Available: http : / / www . havsvind . org / wp - content/uploads/2017/03/S2-Denmark-1.pdf. [8] “Energiaftale; Danish Energy Agreement,” Tech. Rep., 2018. [9] Energistyrelsen, “Energistatistik 2017,” Tech. Rep., 2017. [Online]. Available: https://ens.dk/sites/ens.dk/files/Statistik/pub2017dk.pdf. [10] 4COffshore, Anholt Offshore Wind Farm. [Online]. Available: https://www. 4coffshore.com/windfarms/anholt-denmark-dk13.html (visited on Octo- ber 1, 2018). [11] ——, Horns Rev 3 Offshore Wind Farm. [Online]. Available: https://www. 4coffshore.com/windfarms/horns-rev-3-denmark-dk19.html (visited on October 1, 2018). [12] ——, Kriegers Flak Offshore Wind Farm. [Online]. Available: https://www. 4coffshore.com/windfarms/kriegers- flak- denmark- dk37.html (visited on October 1, 2018). 72 References

[13] Vattenfall, Om Vesterhav Nord. [Online]. Available: https://corporate.vattenfall. dk/vores- vindmoller- i- danmark/vindprojekter/vesterhav- nord/om- vesterhav-nord/ (visited on November 1, 2018). [14] ——, Om Vesterhav Syd. [Online]. Available: https://corporate.vattenfall. dk / vores - vindmoller - i - danmark / vindprojekter / vesterhav - syd / om - vesterhav-syd/ (visited on November 1, 2018). [15] BVG Associate and InnoEnergy, “Future renewable energy costs: Offshore wind,” Tech. Rep., 2017. [16] B. Möller, L. Hong, R. Lonsing, and F. Hvelplund, “Evaluation of offshore wind resources by scale of development,” Energy, volume 48, 2012. [Online]. Available: http://dx.doi.org/10.1016/j.energy.2012.01.029. [17] G. Hundleby and K. Freeman, “WindEurope: Unleashing Europe’s offshore wind potential - A new resource assessment,” Tech. Rep., 2017. [18] K. Veum, L. Cameron, D. Huertas Hernando, and M. Korpaas, “Roadmap to the deployment of offshore wind energy in the Central and Southern NorthSea (2020-2030),” Tech. Rep., 2011. [Online]. Available: http://www.windspeed. eu / media / publications / WINDSPEED{\ _ }Roadmap{\ _ }110719{\ _ }final . pdf{\ % }5Cnhttp : / / scholar . google . com / scholar ? hl = en{\ & }btnG = Search{\&}q=intitle:Roadmap+to+the+deployment+of+offshore+wind+ energy+in+the+Central+and+Southern+North+Sea+(2020-2030){\#}0. [19] D. Bogdanov and C. Breyer, “Integrating the Excellent Wind Resources in North- West Eurasia for a Sustainable Energy Supply in Europe,” 15th Wind Integration Workshop, 2016. [20] D. Hdidouan and I. Staffell, “The impact of climate change on the levelised cost of wind energy,” Renewable Energy, 2017. [Online]. Available: http://dx.doi. org/10.1016/j.renene.2016.09.003. [21] C. Mattar and M. C. Guzmán-Ibarra, “A techno-economic assessment of off- shore wind energy in Chile,” Energy, 2017. [Online]. Available: https://www. sciencedirect.com/science/article/pii/S0360544217308551. [22] P. Beiter, W. Musial, A. Smith, L. Kilcher, R. Damiani, M. Maness, S. Sirnivas, T. Stehly, V. Gevorgian, M. Mooney, and G. Scott, “A Spatial-Economic Cost- Reduction Pathway Analysis for U.S. Offshore Wind Energy Development from 2015–2030,” number September, 2016. [Online]. Available: https://www.nrel. gov/docs/fy16osti/66579.pdf. [23] P. Beiter, W. Musial, L. Kilcher, M. Maness, A. Smith, P. Beiter, W. Musial, L. Kilcher, M. Maness, and A. Smith, “An Assessment of the Economic Potential of Offshore Wind in the United States from 2015 to 2030,” 2017. [24] A. D. Mills, D. Millstein, S. Jeong, L. Lavin, R. Wiser, and M. Bolinger, “Esti- mating the Value of Offshore Wind Along the United States ’ Eastern Coast,” page 21, 2018. References 73

[25] Danish Ministry of Energy Utilities and Climate, Kriegers Flak - the largest offshore wind farm in the Baltic. Sea [Online]. Available: http://efkm.dk/ media/7878/factsheet-kriegers-flak.pdf (visited on January 20, 2019). [26] V. Negro, J. S. López-Gutiérrez, M. D. Esteban, P. Alberdi, M. Imaz, and J. M. Serraclara, “Monopiles in offshore wind: Preliminary estimate of main dimensions,” Ocean Engineering, 2017. [Online]. Available: http://dx.doi. org/10.1016/j.oceaneng.2017.02.011. [27] T. Buhl and A. Natarajan, “Level 0 cost models of offshore substructure -A simple cost model including water depth,” DTU - Department of Wind Energy, Tech. Rep., 2015, page 13. [28] H. Smail, R. Alkama, and A. Medjdoub, “Optimal design of the electric connec- tion of a wind farm,” Energy, 2018. [Online]. Available: https://linkinghub. elsevier.com/retrieve/pii/S0360544218319960. [29] EMD International A/S, windPRO: Siemens wind turbine SWT 3.6-120. [On- line]. Available: https://www.emd.dk/windpro/. [30] GE Renewable Energy, Haliade-X Offshore Wind Turbine Platform. [Online]. Available: https://www.ge.com/renewableenergy/wind-energy/turbines/ haliade-x-offshore-turbine (visited on December 3, 2018). [31] C. Bak, F. Zahle, R. Bitsche, T. Kim, A. Yde, L. C. Henriksen, A. Natarajan, and M. Hansen, “Description of the DTU 10 MW Reference Wind Turbine,” DTU Wind Energy, Tech. Rep., 2013. [32] A. Bech Abrahamsen, D. Liu, and H. Polinder, “Direct drive superconducting generators for INNWIND . EU wind turbines Document information,” DTU Wind Energy, Tech. Rep. 308974, pages 1–137. [33] MHI Vestas Offshore Wind, The V164-8.0 MW Turbine. [Online]. Available: http : / / www . mhivestasoffshore . com / innovations/ (visited on Sep. 13, 2018). [34] DTU, Global Wind Atlas 2.3, 2018. [Online]. Available: https://globalwindatlas. info/. [35] A. Bech Abrahamsen, “DTU Course 47202: Introduction to future energy, Lec- ture: Introduction to wind energy,” 2018. [36] DTU Wind Energy, WAsP. [Online]. Available: http://www.wasp.dk/. [37] EMODnet Bathymetry portal, Bathymetry Viewing and Downloading Service. [Online]. Available: http://portal.emodnet- bathymetry.eu/ (visited on Sep. 2, 2018). [38] International Renewable Energy Agency (IRENA), Renewable Power Genera- tion Costs in 2017. 2018. arXiv: arXiv:1011.1669v3. [39] Danish Energy Agency, “Note on technology costs for offshore wind farms and the background for updating CAPEX and OPEX in the technology catalogue datasheets,” pages 1–11, 2018. 74 References

[40] SteelBenchmarker, “Price History,” Tech. Rep., 2018. [41] H. Kooijman, M. de Noord, C. Volkers, L. Machielse, F. Hagg, P. Eecen, J. Pierik, and S. Herman, “Cost and Potential of Offshore Wind Energy on the Dutch part of the North Sea,” 2008. [42] Douglas Westwood, “Offshore Wind Assessment For Norway - Final Report,” Tech. Rep., 2010. [43] Danish Energy Agency, “Technology data for energy plants. Generation of elec- tricity and district heating, and energy carrier generation and conversion,” Tech. Rep., 2012. [Online]. Available: https://www.energinet. dk/SiteCollectionDocuments/Danskedokumenter/Forskning/Technology{\_ }data{\_}for{\_}energy{\_}plants.pdf. [44] P. Beiter, W. Musial, A. Smith, L. Kilcher, R. Damiani, M. Maness, S. Sirni- vas, T. Stehly, V. Gevorgian, M. Mooney, and G. Scott, “A Spatial-Economic Cost- Reduction Pathway Analysis for U.S. Offshore Wind Energy Develop- ment from 2015–2030,” 2016. [Online]. Available: https://www.nrel.gov/ docs/fy16osti/66579.pdf. [45] R. Lacal-Arántegui, J. M. Yusta, and J. A. Domínguez-Navarro, “Offshore wind installation: Analysing the evidence behind improvements in installation time,” Renewable and Sustainable Energy Reviews, 2018. [46] Ballast Nedam, “Optimal integrated combination of foundation concept and installation method,” Tech. Rep., 2009. [47] Y Dalgic, I Lazakis, and O Turan, “Vessel charter rate estimation for offshore wind O&M activities,” Developments in Maritime Transportation and Exploita- tion of Sea Resources, 2013. [48] T. Schlemmer and L. Greedy, “66 kV Systems for Offshore Wind Farms,” Tech. Rep., 2015. [49] A. G. Gonzalez-Rodriguez, “Review of offshore wind farm cost components,” Energy for Sustainable Development, 2016. [50] Energinet, Transmission system data. [Online]. Available: https://en.energinet. dk/Electricity/Energy-data/System-data (visited on Sep. 16, 2018). [51] X. Xiang, M. M. C. Merlin, and T. C. Green, “Cost Analysis and Comparison of HVAC , LFAC and HVDC for Offshore Wind Power Connection,” IET 12th International Conference on AC andDC Transmission, 2016. [52] NREL, Offshore Wind Power Plants. [Online]. Available: https://atb.nrel. gov/electricity/2017/index.html?t=ow (visited on October 29, 2018). [53] W Musial and B Ram, “Large-Scale Offshore Wind Power in the United States: Assessment of Opportunities and Barriers, NREL (National Renewable Energy Laboratory),” Tech. Rep. September, 2010. [Online]. Available: http://www. osti.gov/bridge:{\%}5Cnhttp://www.ntis.gov/ordering.htm. References 75

[54] Ernst & Young, “Cost of and financial support for offshore wind Private and confidential Executive summary Background,” Tech. Rep. April, 2009. [Online]. Available: http://webarchive.nationalarchives.gov.uk/+/http:/www. berr.gov.uk/files/file51142.pdf. [55] W. Engels, T. Obdam, and F. Savenije, “Current developments in wind - 2009; Going to great lengths to improve wind energy,” 2009. [56] R. H. Wiser, K. Jenni, J. Seel, E. Baker, M. M. Hand, E. Lantz, and A. Smith, “Forecasting Wind Energy Costs and Cost Drivers: The Views of the World’s Leading Experts,” Lbnl-1005717, 2016. [Online]. Available: https:// escholarship.org/uc/item/0s43r9w4. [57] DTU Wind Energy, Fuga: Wake model for offshore wind farms. [Online]. Avail- able: http://www.wasp.dk/fuga. [58] N. G. Mortensen and The WAsP team, “DTU course 46200: Class: Wind re- source mapping and wind farm modelling,” 2017. [59] L. Kitzing, D. Møller-Sneum, and R. Bramstoft, “DTU course 42004: Feasibility studies of energy projects, Lecture note 10: Cost of capital,” 2017. [60] Vindmølleindstrien, Kystnære havvindmøller fortsat billigst. [Online]. Available: http://eolienne.org/da/aktuelt/aktuelt{\_}i{\_}vindmoelleindustrien/ news{\_}q4{\_}2016/kystnaere{\_}havvindmoeller{\_}fortsat{\_}billigst. html (visited on October 10, 2018). [61] Siemens-Gamesa, SG 8.0-167 DD - Offshore wind turbine. [Online]. Available: https : / / www . siemensgamesa . com / en - int / products - and - services / offshore/wind-turbine-sg-8-0-167-dd (visited on October 1, 2018). [62] T. Freyman and T. Tran, “Grant Thornton: Renewable energy discount rate survey results – 2017,” 2018. [63] L. Schröder, “Screening potential areas and techno-economic assessment of wind farms for the planning of Danish offshore wind energy exploitation,” Tech. Rep., 2016. [64] T. Stehly, D. Heimiller, and G. Scott, “NREL: 2016 Cost of Wind Energy Re- view,” Tech. Rep., 2016. [Online]. Available: https://www.nrel.gov/docs/ fy18osti/70363.pdf. [65] European Commission, EU Emissions Trading System (EU ETS). [Online]. Available: https : / / ec . europa . eu / clima / policies / ets{\ _ }en (visited on January 17, 2019).