The influence of offshore wind energy on the electricity market
C. Kleinschmidt (KEMA) N. Moldovan (KEMA) H. Cleijne (KEMA) F. Verheij (KEMA)
(We@Sea project 2005-021)
50680533-Consulting 07-0048
The influence of offshore wind energy on the electricity market
Arnhem, 21 December 2006 Authors: Chris Kleinschmidt, Natalia Moldovan, Hans Cleijne, Frits Verheij KEMA Consulting
KEMA Nederland B.V. Utrechtseweg 310, 6812 AR Arnhem P.O. Box 9035, 6800 ET Arnhem The Netherlands T +31 26 3 56 91 11 F +31 26 3 89 24 77
[email protected] www.kema.com Registered Arnhem 09080262
© KEMA Nederland B.V., Arnhem, the Netherlands. All rights reserved.
It is prohibited to change any and all versions of this document in any manner whatsoever, including but not limited to dividing it into parts. In case of a conflict between the electronic version (e.g. PDF file) and the original paper version provided by KEMA, the latter will prevail. KEMA Nederland B.V. and/or its associated companies disclaim liability for any direct, indirect, consequential or incidental damages that may result from the use of the information or data, or from the inability to use the information or data contained in this document. The contents of this report may only be transmitted to third parties in its entirety and provided with the copyright notice, prohibition to change, electronic versions’ validity notice and disclaimer.
-3-
50680533-Consulting 07-0048
ACKNOWLEDGEMENT The authors would like to thank all persons within KEMA and of other companies who have contributed to this study. We gratefully acknowledge Jan van den Bor of Nuon Energy Trade & Wholesale, Mathieu Kortenoever of E-Connection and Huub den Rooijen of Shell WindEnergy who provided important input to this study during our interviews, workshops and other contact moments. KEMA owes many thanks to Bernd Tersteegen who did his thesis in the second half of 2005 at KEMA and who started the inventory of all relevant reports and other documents related to our study.
-4-
50680533-Consulting 07-0048
CONTENTS Page Executive Summary .............................................................................................................. 6 1
Introduction .........................................................................................................16
2
Key issues derived from literature inventory........................................................18
2.1 2.2 2.3 2.3.1 2.3.2 2.3.3 2.3.4 2.3.5 2.4 2.5
Context of the inventory ......................................................................................18 The Dutch generation system .............................................................................19 The Dutch market structure and price mechanisms.............................................21 TenneT Balancing market ...................................................................................21 APX (Amsterdam Power Exchange) ...................................................................23 Over The Counter market....................................................................................24 ENDEX Futures Exchange..................................................................................24 Market regulation ................................................................................................25 Wind power characteristics .................................................................................27 Valuing power plants...........................................................................................32
3 3.1 3.2 3.2.1 3.2.2 3.2.3 3.2.4 3.3
Wind power on the wholesale market..................................................................37 Factors that influence the electricity wholesale price ...........................................37 The impact of wind power on the wholesale market ............................................39 Effect on the APX-spot market ............................................................................39 Influence on future markets (ENDEX) .................................................................40 Influence on the imbalance market......................................................................41 Effect on fuel prices.............................................................................................42 Correlation between wind power and electricity prices ........................................42
4
The value of wind power .....................................................................................44
4.1 4.2 4.3 4.4
Overview of value and cost components .............................................................45 Value components ..............................................................................................47 Cost components ................................................................................................51 Discussion and preliminary results ......................................................................54
5 5.1 5.2 5.3
Methodology of a model......................................................................................57 Perspective of the PRP .......................................................................................57 Scope of the simulation.......................................................................................58 ProSym and Symbad ..........................................................................................60
-5-
50680533-Consulting 07-0048
6
Scenarios of the development of the electricity market........................................63
7
Conclusions ........................................................................................................65
References...........................................................................................................................67 Appendix I Definitions of key terms ..................................................................................71 Appendix II Quantitative correlation analyses....................................................................80 Appendix III Calculation methodology capacity credit.........................................................86 Appendix IV Technical data of a Generation unit - PROSYM input data .............................88 Appendix V Calculating the imbalance costs resulting from a forecast error ......................89 Appendix VI Production Simulation Tool - ProSym .............................................................91 Appendix VII SYMBAD ........................................................................................................93
-6-
50680533-Consulting 07-0048
EXECUTIVE SUMMARY Both electricity price and energy yield of wind farms vary in time resulting in uncertainties about income and value of these wind farms and, thus, in economic and business risks. The production volume of wind power has increased over the last decades and will further increase in the next decades. In The Netherlands this growth will mainly be caused by large offshore wind farms. Numbers of 6 GW and more installed wind power in 2020-2030 are used in government scenarios and by other organisations. In general this increase in wind power – absolute (number of MW’s of installed power) and relative (penetration level of wind power) – will most probably cause an increase in uncertainties for wind farm owners, investors and Programme Responsible Parties (PRPs). But what are the risks, especially the financial risks, of wind energy for those parties? Our study aimed at creating more transparency in the costs and value of (offshore) wind power in the complex electricity market mechanisms and, thus, creating a more clear picture of the financial risks of offshore wind energy for developers, wind farm owners, investors, traders and PRPs.
Key issues derived from literature study In the first phase of the present study we have carried out an extensive information search to get an overview of relevant information needed to comprehend the Dutch electricity system. This overview includes characteristics of a wind energy plant, characteristics of the Dutch energy production mix, structure of the wholesale electricity market and more specifically the imbalance settlement mechanism. Furthermore, international integration studies were included in this overview to compare their findings with our study and to learn from their perspectives and methods. The findings of this inventory study laid the foundation for our economic assessment of offshore wind power integration. We have summarized the relevant examples of knowledge that created the context of this assessment. Electricity market - Because of its long-term unpredictability, all electricity generated by a wind farm will be sold on the day-ahead electricity market. Wind power is first in the market merit order. This is the result of negative marginal production costs when MEP subsidies are taken into account (MEP is regarded as opportunity costs);
-7-
-
-
-
-
-
1
50680533-Consulting 07-0048
Utilities differ in their strategies to compensate for wind power, depending on the content of their portfolio and their (wind) power purchase agreements. Some utilities will selfprovide balancing power using their own conventional generation units, other utilities let TenneT procure balance power and allocate the costs to the wind farm owner. In a perfectly functioning market the result will be the same; The relatively large number of gas-fired plants in The Netherlands can be regarded as an advantage when installing large volumes of wind power as these plants are able to quickly respond to imbalance situations, e.g. caused by forecast errors of wind power output. Hydro power would even be better but there are only a few very small hydro power plants available in The Netherlands. However, during night hours the operation capacity comprises mostly coal-fired and nuclear power plants which can provide 60 MW/min ramp up or down power (We@Sea, 2006). When 1000 MW offshore wind capacity becomes unavailable within 10 minutes, because of a storm, the system will not be able to compensate wind power’s imbalance; The electricity market in The Netherlands consists of several market places. More than 85% of the annual traded electricity volume takes place on the bilateral market. The APX (day-ahead or spot market) might be the most well-known although its volume is no more than some 15% of the electricity demand (16.0 TWh in 2005). A large part of the traded energy volume on the APX comprises imported energy (5 TWh yearly, i.e. some 30% of APX’s traded volume). Other marketplaces are OTC and ENDEX on which electricity can be traded between one hour up to 3 years. Last but not least the imbalance market1, operated by TenneT, is used to balance the real-time deviation between scheduled and actual energy production and consumption. Although the latter “market” is small, some 3% of the wholesale market, the influence on the wholesale price can be considerable. The Dutch imbalance market is regulated on a two-price model. The TSO charges imbalance costs or pays PRPs for their overall imbalance energy, without specifying the imbalance costs to individual producers such as wind farm owners. At present a fixed fee (premium) to cover the costs of the imbalance expected to be caused by a wind farm is part of a PPA contract between a wind farm owner and a PRP. The method used by the PRPs to determine the wind power’s imbalance costs is not transparent. The price settlement on the Dutch imbalance market results in a dissymmetric price structure, where the average ramp-up prices are almost two times higher than the average ramp-down prices. For this reason most of the time the system is long. By being long PRPs avoid higher imbalance costs.
or Balancing market, known in the Netherlands as the Imbalance Price System
-8-
50680533-Consulting 07-0048
Wind power characteristics - Both wind turbine technology and the accuracy of predicted wind power continues to improve, which reduces wind power’s impact on system imbalance. Most modern wind turbines use variable speed and pitch control increasing the controllability of wind turbines (farms). Recently the wind turbine manufacturers of these turbines have also improved the control system of their machines enabling them to gradually reduce the power level above the cut-out wind speeds (approximately 24 m/s); - Most studies agree that hourly forecast errors in the day-ahead programmes will remain < 14% for 90% of the time. The forecast error significantly increases when the prediction window becomes larger, to 50% for a prediction window of 36 hours; - Currently, wind power producers are forced to predict the wind power output of their wind farms 12 to 36 hours in advance as practically no intra-day market exists in The Netherlands. As the APX started an intra-day market as of 14th September 2006, the possibilities to get better wind power balancing will improve significantly. Depending on the liquidity of the intra-day market, this can result in lower imbalance costs. If traded energy volumes are small and only few parties participate, the effect on wind power’s imbalance costs will be minimal; - Wind power’s variability on an hourly time-scale will affect the so-called secondary control mechanism of the electricity system, which comprises the load following capabilities of the operational production units; - Wind power generation shows a daily periodical behaviour, most pronounced in summer, with the highest generation in the late afternoon and the lowest in the early morning. Both seasonal and daily periodicities more or less coincide with the normal load fluctuations in a power system, with the highest loads in winter and in the late afternoon. This would give reasons to believe that wind power capacity represents a certain load following value; - Many studies conclude that wind power will not effect short-term (minute-to-minute timescale) system balancing. We found two reasons for this. First the wind power fluctuations are small compared to other disturbances of the power balance (load variations). This might change when significant volumes of wind are installed, although these minute-to-minute fluctuations will even out over a large number of wind turbines especially if these wind turbines are greographically spread out. Secondly, the shortterm wind speed variations are smoothed out by variable speed technology, the inertia of large rotors and the variations in power output of individual wind turbines within a wind farm; - Technical availability of modern wind turbines is high, several sources show figures of some 98%. For offshore wind farms these figures have not been demonstrated yet. The availability of wind however is uncertain. As this primary energy source cannot be stored
-9-
50680533-Consulting 07-0048
and is not fully predictable, wind power has a negative effect on the short-term system security. System integration and capacity credit2 - The value of a power plant increases with its ability to generate power when it is needed, as this increases its contribution to system reliability (capacity value). Every MW of generation capacity that is added to the system reduces the risk of a loss of load situation; - As the capacity of power plants differs in quality a distinction is made between firm and non-firm capacity. Power plants with firm capacity can control their production level and have a high availability. Power plants with non-firm capacity, such as wind power, have limited control over their production level and an uncertain availability; - Besides it avoids fuel costs, every additional MW of wind power reduces the loss of load of the system. The capacity credit of wind power is based on the effective load carrying capability of a wind portfolio. The effective load carrying capability (ELCC) is a percentage of the total wind portfolio that results in the same loss of load probability (LOLP) in a generating portfolio compared to an equal amount of firm capacity; - Utilities are not awarded a capacity value in addition to the electricity price. As a consequence wind power is paid the same capacity value as conventional power plants through the wholesale electricity price. As wind power does not have the same capacity credit as conventional power, a risk premium is subtracted from the wind power contract price; - The capacity credit will increase with the increase of wind power capacity, but not proportionally. The additional capacity credit will become smaller with every additional installed MW of wind energy. Seasonal effects influence the capacity credit; - The capacity credits3 of wind power can be used as hedging instrument for the fuel price variations and price volatility over time. Although, spark spread contracts and options are common hedging tools, long-term traders can speculate on the capacity credit of the national wind power portfolio (this issue is still controversial and should be further analysed); - When wind penetration increases the average imbalance volume during each PTU (15 minutes period) will also rise, which results in a higher average price for reserve and regulation power.
2
the effective load carrying capability of a wind portfolio as percentage of the total wind portfolio which will be available at the same reliability as firm capacity 3 the effective load carrying capability of a wind portfolio as percentage of the total wind portfolio which will be available at the same reliability as firm capacity
-10-
50680533-Consulting 07-0048
The effect of wind on the wholesale market In case large amounts of (subsidized) wind energy are fed into the electrical grid and, thus, are available on the electricity market wind power will affect the price level on the wholesale market. Most directly this will be visible on the spot market (APX) and the imbalance “market”, but it will also trickle through to future/forward prices and electricity options. We have made qualitatively analyses to derive the influencing factors on the wholesale market prices. We continued with quantitative analyses using correlation techniques on data available from June 2004 to May 2005 to study the influence of wind power on the wholesale market. The main data has been obtained by KEMA in a former We@Sea study on system integration and balance preservation of 6,000 MW offshore wind power. Only a few per cent of the variation in the APX price can be explained by variations in wind speed forecasts. However, in January and February 2005 between hour 0 and 6 a.m. on non-working days the hourly correlations of the APX prices and the on-shore wind speed forecast show relatively high negative values (-0,86 < R < -0,45). In one of the data series we also found high negative correlations during the same nightly hours in May 2005. During these specific hours occasionally 20-75% of the changes in the APX price is caused by wind power. This can be partly explained by the coexistence of relatively high wind speeds and low APX prices during nightly hours. Furthermore, during these hours the system switched over from “night load” to “day load” which means that the APX price was settled at the steep part of the market merit order. At those moments wind power production could lead to relatively large price reductions. To conclude, only during very few hours per year wind speed forecasts – indicating the wind power production – have a significant price reducing effect on the APX. During all other hours this effect is hardly recognizable. The correlation analyses have also been used to explore the relationship between wind speed forecast error and system imbalance. The first conclusion on these analyses is that a significant positive correlation is shown occasionally between a short system balance – when electricity consumption is higher than the electricity consumption Programme Responsibility Parties have been forecasted – and a negative wind speed forecast error (more wind power has been produced than forecasted). The frequency of occurrence of values R2 > 0.2 is about 3.3% of the time.
-11-
50680533-Consulting 07-0048
Significant positive correlations also are shown occasionally between a long system balance – when electricity consumption is lower than the electricity consumption Programme Responsibility Parties been have forecasted – and the positive wind speed forecast error (more wind power was produced than forecasted). For these situations the frequency of occurrence of values R2 > 0.2 is about 3.8% of the time. Most of the above mentioned correlations have been found in sub series of nightly hours during non-working days, i.e. during hours with relatively low electricity demand. The overall conclusion is that, in general, the monthly and yearly correlations show that the wind power portfolio installed in 2004-2005 has almost no effect on the APX price and little effect on the TenneT imbalance market. In order to draw more conclusions data series covering a longer period (a couple of years) and/or more detailed analyses of the existing data using further segmentation, as for different wind speeds and different APX price levels, are required.
The value of wind power The findings of the inventory study have also been used to list all factors that can potentially influence the price of electricity produced by an (offshore) wind farm. After discussion with our partners E-Connection, Nuon and Shell the list has been finalized. The influencing factors are: wholesale market, quality and accuracy of wind forecasts, variability of wind speeds, availability of a wind farm, (rate of) wind penetration, fossil production mix, system balance (including reserve capacity), load, oil/gas prices, customer preferences and legislation and regulation. The main component that determines the price for wind power is the market value4 (based on the wholesale electricity price). However, it seems that not all effects of wind power are included in this market value. In order to approach (estimate) the real value of wind power we have distinguished (added) value and cost components5. The (added) value components comprise: the image of green energy, wind pattern value, geographical spread, fuel diversification, capacity value and
4 5
revenues for physical produced energy to be addressed as value components
-12-
50680533-Consulting 07-0048
down regulating power. The MEP is an important price component of wind power, but should not be regarded as a value component as this subsidy is determined by the difference between wind power’s market value and its production costs. Thereby, the MEP subsidy is an output variable, instead of an input variable. The cost components are: short-term imbalance cost, long-term risk and administrative costs. Except for the latter, these cost components are mainly dependent on the volume and price risk of wind power. To enable a simple overview of which factors influence the individual wind power’s price components we presented the relationships in one matrix. In the table below the price components that determine the wind power price are given including the methodology as well as the financial instrument needed to derive the price. In the last column the tools we used or will use in the next phase are shown.
-13-
50680533-Consulting 07-0048
Table: Overview of the wind power’s value and cost components analysis Price components of wind power
Methodology
Derivative
Tools
Market value
Revenues for produced physical energy
Value of day-ahead contract (APX)
Excel / ProSym and Symbad
Subsidy (MEP)
Fixed value during 10 years, strongly related to market value
-
Not a perspective of this study
Image PRP (more customers)
Interviews, literature, number of switches
-
Not been studied in detail in this project
Wind pattern value
Calculate value of wind production forecast on APX. (hourly granularity)
Value of day-ahead contract (APX)
Excel / ProSym
Geographical spread
Correlation between wind parks
-
Not been studied in detail in this project
Fuel diversification
Value premium of long-term hedging products; Simulation of production units
Premium for call option oil/gas (fuel market)
None / ProSym and / or Excel
Plant Margin Effect (capacity value)
Simulation of production units
Firm-capacity for 10 years
None / ProSym and Symbad
Down regulating power
Deterministic (15 minutes granularity)
Ramp-down price
Excel / ProSym and Excel
Short-term imbalance cost
Quantify the forecast7 error of wind energy
Imbalance product
Excel / Prosym and Symbad
Long term risks
Value portfolio risk (arbitration between ENDEX and APX)
Value the premium for mandatory trading in order to balance the portfolio due to wind variation over time
None / Prosym, Excel and Simbad
Administrative costs
Value program handling costs
Program handling
-
(this study/ next phase)
Value components 6
Cost components
6 7
is the image a PRP / utility with wind in its portfolio has in the eyes of clients forecast error = actual (measured) - forecast
-14-
50680533-Consulting 07-0048
The next step was to explore the relationships between the price influencing factors and the wind power’s price components both qualitatively through interviews and quantitatively through calculations on existing data series including the aforementioned correlation analyses. The main components are the market value and the short-term imbalance cost. In order to get a first ‘feeling’ over these values we used historical APX / TenneT prices and wind speed data (forecast and actual) for every hour / PTU (15 minutes period) during the period of 1st June 2004 – 31st May 2005. We applied these data to a fictitious 100 MW wind farm, the one time for an onshore location and the other for an offshore location. We assumed that the owner (PRP, fully exposed to imbalance market) is operating this wind farm without any other asset and therefore is not able to compensate for the imbalance caused by the farm. We were able to quantify the most important determinants in the wind power price by using data series of wind power production at 3 different locations in The Netherlands. The average market value of the electricity produced by these wind farms ranges between 34 and 39 Euro/MWh. When we consider the down regulation value of these wind farms in a MEP regulated regime the values ranged between 2.4 and 3.5 Euro/MWh. Without the MEP, the opportunity costs of missed production income will reduce, which resulted in a down regulation value between 4.4 and 5.5 Euro/MWh. The short-term imbalance costs are the lowest for an offshore wind park with a weighted average of 2.2 Euro/MWh, while the short-term imbalance costs for an onshore wind park resulted in a weighted average between 5.6 and 8.8 Euro/MWh. For all data series the onshore imbalance costs represented approximately 20% of total wind production income on the APX. For the offshore wind farm used in our case study the imbalance costs represents only 6% of the total income when the electricity produced would have been sold on the APX.
The next phase In the next, 2nd phase, of this study the aim is to develop a model using ProSym and Symbad to quantify all value and cost components of wind energy for a given market configuration based on simulations and their resulting spot price and imbalance price. In the present study we have drafted the methodology of this model. One of the starting points is that the calculations will be carried out from a market perspective. The wind farm owner might be a (small) company that only owns one or several (offshore) wind farms, a
-15-
50680533-Consulting 07-0048
company that also owns other assets like a stand alone gas turbine or this company can be a utility with a variety of production units. The strategy of the company can be to maximise its profits or to reduce its risks. Another starting point is the assumption of a perfect functioning electricity market, i.e. price of electricity is determined by the intersection point of demand and supply curves as production units are running on short term marginal costs. Some of the input variables of the model will be: production plants and their technical data, hourly import and export volumes and capacities, hourly demand, regulations changes, hourly wind speed profile (based on historical data) and wind capacity penetration. KEMA will use the model to calculate the value and cost components for years 2005, 2010 and 2020 for several scenarios. These scenarios will be derived from studies performed by CBS, TenneT, UCTE and KEMA (other We@Sea research study) to be consistent with the basic assumptions used in former studies. The input data for the basic scenario is given in this report.
-16-
1
50680533-Consulting 07-0048
INTRODUCTION
The electricity production of an (offshore) wind farm and the price the wind farm owner receives for the produced electricity define the revenues for the owner. The costs of a wind farm (investment costs, operational costs and other costs) can be estimated within reasonable ranges. However, electricity price and production volume vary as a function of time resulting in uncertainties about this yield. This creates a financial risk for the wind farm owner, the company buying the produced electricity, the electricity traders and the financers. Each of them will try to reduce its own risk as much as possible thus indirectly increasing the risks of the other parties involved. The scope of this study and model is to value each particular risk (by defining and valuating wind price components) and allocate each particular risk to the party who has or should have the instruments to manage it. As the volume of onshore and especially offshore wind will grow significantly in the coming years the total size of the risks most probably will increase. Dutch utilities indicate that current knowledge about the physical and financial risks, concerning the integration of a substantial amount of offshore wind energy, is still insufficient. There is no clear understanding of the risks associated with integrating a ‘power plant’ which has lower flexibility, lower predictability and higher variability than conventional power plants. As long as those risks remain unidentified and the logical framework to quantify those risks is missing, the financing, insuring and sales of wind power will be limited by uncertainties. Earlier work of KEMA and some partners in the We@Sea consortium has already resulted in a model to calculate the physical effects of large-scale wind penetration in the Dutch electricity system. However, to take this work a step further, a translation of physical effects into commercial implications is needed. This is a challenging task, because the cost and value of wind energy are dependent on complex market mechanisms, which have a limited transparency and many controversial aspects such as allocation issues. In the present study we have mainly been focusing on building up a logical framework that will serve as validation for quantifying the commercial implications of wind energy. Aim is to derive all value and cost components which should be included in the price of offshore wind energy and to determine which factors influence those components. We have carried out statistical analyses to derive the correlation between a few of these influencing factors and some of the value and cost components. This study concentrates on the commercial items of the integration of wind power in the Dutch electricity market. The technical aspects like grid integration are no part of this study.
-17-
50680533-Consulting 07-0048
Outline of the report An overview has been made of all relevant information needed to comprehend the Dutch electricity system. A very extensive summary of our inventory of existing knowledge is written in a separate report. Our key findings are given in chapter 2 of this report. Chapter 3 provides a brief description of the effect of wind power in the wholesale market. Amongst others it shows an overview of the factors that influence the wholesale price of electricity. This is an important input for chapter 4 in which we explain how one can derive the value of wind power. It starts with an overview of factors influencing the price of electricity produced by a wind farm followed by an analysis of all value and cost components of this price. In chapter 5 the methodology of a model is described in order to quantify abovementioned value and cost components. Developing this model will be part of phase 2 of our study. In this study we will perform the analyses from a market perspective. Scenarios of the development of the electricity market and of the development of (offshore) wind energy in the next decade are given in chapter 6. More detailed figures will be given in phase 2. This report ends with our main conclusions in chapter 7.
-18-
2
50680533-Consulting 07-0048
KEY ISSUES DERIVED FROM LITERATURE INVENTORY
In this chapter the aspects of the Dutch electricity system relevant for this study will be explained. This comprises key issues about the generation system (2.2), the electricity market (2.3) and wind power characteristics (2.4). In the first section (2.1) the context in which we will assess the impact of (offshore) wind power integration on the (Dutch) electricity market is described.
2.1
Context of the inventory
At any moment the production of electricity has to be balanced with the demand (load) of electricity. On real time, the Dutch independent Transmission System Operator (TSO) TenneT is responsible for this balancing. The other main players in this area are PRPs and energy producers. If one of these players causes imbalance, this will be settled with TenneT (the imbalance costs). In order to reduce imbalance (match between buy/supply and sell/demand electricity volumes) – and imbalance costs – electricity can be traded between parties. This occurs in different time frames up to almost one hour before real time and on several marketplaces, see section 2.3 Depending on his fuel mix portfolio mix, his contracts portfolio and his strategy, a market party can decide the point in time when he will trade and / or transfer his portfolio to a PRP. Utilities manage short-term and long-term positions (up to three years), e.g. long-lasting contracts, and use short-term forecasts of the loads during the last days before the program time unit (PTU) in question. Then they use these forecasts to adjust their long-term positions for matching purposes. One day ahead at noon the PRPs have to submit their Electricity-program for all PTUs of the next day. Intraday producers dispatch their power plants to meet the requested loads. Different power plants are in place to produce the required electricity. Gas-powered plants are flexible, coal and especially nuclear power plants are less flexible. The latter ones are mainly used to cover the base loads. The Dutch production portfolio has changed over the years. Global warming, rising gas and oil prices and an increasing import of electricity influences governmental policy on energy. These themes also affect social and environmental awareness which resulted in a growing demand for greener energy over the last years. Wind power is one of the main contributors to
-19-
50680533-Consulting 07-0048
greener energy, which will even be more important as offshore wind farms are foreseen to produce large volumes of ‘green electricity’ in the next decades. Due to the unknown fluctuations in wind speed, wind farms are often regarded as uncontrollable and inflexible power plants. At least unexpected power variations or deviations between forecasted and actual wind electricity volumes have to be counterbalanced by other, flexible (relatively expensive) power plants causing a cost component for wind power. These and other cost and value components of wind power are sometimes unknown or at least not fully transparent at the moment which makes it hardly possible to depict the right price for electricity generated by wind farms. In this project we aim to identify cost and value components of wind power and define methodologies to quantify these components in the Dutch power market. The first question thus is to get clear information on the relevant elements of the generation system (2.2), the way the electricity market is functioning (2.3) and on the main wind power characteristics (2.4). In this study we will concentrate on the Dutch situation although we also studied markets in other countries for comparison purposes.
2.2
The Dutch generation system
In order to assess the impact of wind power we need to consider the composition of the generation system. Each generation type has its own dependencies and limitations, which can be described by the variability, controllability and (technical) availability. These three aspects determine the ability of a power plant to fulfil both the energy production task and a balancing task. Especially, the balancing capabilities of conventional production units are crucial for offshore wind power integration. Table 2.1 shows figures about the key characteristics of the large-scale centralised production capacity (>60 MW). The decentralised generation capacity is not included in this table because it makes a minor contribution to balancing the system and is mostly used for local electricity demand in the agricultural sector, district heating or in the industrial sector.
-20-
50680533-Consulting 07-0048
Table 2.1: The Dutch generation system Type of plant
Installed capacity (MW)
Percentage of total capacity (%)
Load response (run-up minutes)
Average ramp speed (MW/ min)
Availability
Gas-fired plants
11,190
53%
120-780
80
High
Coal plants
3,930
19%
420
59
High
Nuclear plants
450
2%
>2,000
0
High
IGCC plants
250
1%
420
7,5
High
15,820
75%
Subtotal
Table 2.1 shows that the Dutch generation system is dominated by gas-fired production capacity, which generally has a high flexibility and are therefore able to quickly respond to imbalance situations. There is also a significant amount of coal-fired production capacity, which is expected to grow in the coming years. There is only one nuclear reactor left in the Netherlands. Coal power plants and especially nuclear plants have relatively low ramping speeds. Finally, the Netherlands is one of the first countries, which has installed an IGCC plant. The availability, which determines the long-term balancing capabilities of the system, is relatively high for all conventional production plants. These characteristics of the Dutch generation system have important implications for the integration of wind power, because they represent the technical constraints in balancing wind power’s variability and availability. System stability is more sensitive to wind power fluctuation during nightly hours, because the power production is dominated by base-load production units (nuclear and coal-fired power plants) which have limited ramp capabilities. Table 2.1 makes clear that the aggregated effect of load and wind power fluctuations must not exceed 60 MW/min during nightly hours. Theoretically this would correspond with the shut down of 600 MW of wind capacity in 10 minutes. Such a situation is extreme and very rare and could only occur when concentrated quantities of wind power capacity are upset by a nightly storm. In practice, Dutch wind capacity is not concentrated and most storms will take much longer than 10 minutes to pass The Netherlands (1 to 3 hours is more realistic).
-21-
2.3
50680533-Consulting 07-0048
The Dutch market structure and price mechanisms
In the Netherlands the wholesale electricity price is established on different market places and products. The following market structure can be distinguished: -
OTC market ENDEX market APX market TenneT “market” for imbalance
time periods from one hour up to years time periods from one month until 3 years day-ahead (and intra-day) 15 minutes bids On-line
Bilateral market
ENDEX (OTC)
OTC OTC
time
Imbalance market
APX
t=0
More than 85% of the physical trades on the Dutch wholesale market are predominantly completed on the bilateral market. About 20% of the annual electricity consumption, part of the wholesale market, is imported. The level of cross-border capacity made available to the free market, which also affects most of the above described marketplaces, is supposed to increase in the coming years. In the next paragraphs of this chapter we will shortly present the market places and the priceformation on the Balancing market, the APX market, the ENDEX, the OTC and the bilateral market.
2.3.1
TenneT Balancing market
TenneT operate a Balancing market8 for secondary control power9 and tertiary reserve since 1st January 2001. TenneT buys the secondary reserve from the Balancing Market (the ‘RRV’ 10 market) and sells it to PRPs (Programme Responsible Parties)11 in order to balance their E8
known in the Netherlands as the “onbalansprijssystematiek” secondary reserve 10 Regel en Reserve Vermogen known as Balancing market 9
-22-
50680533-Consulting 07-0048
program12 and the entire system. The balancing mechanism is used for procuring both secondary control and tertiary reserve. Based on the merit order, TenneT selects the least-cost solution for each type of reserve. The prices on both “markets” (Balancing market and Program responsible service) are equal, with exception of the 15 minutes when both up- and downward energy are required. On the Balancing market TenneT buys the secondary reserve in two stages: capacity on annual base13 and energy14 on daily base. Energy bids are valid for a specific 15-minute interval and consist of upward or downward quantity (MWh) and price (€/MWh). The secondary and tertiary reserve has to be offered by production units above 60 MW. Besides, offering is optional for fossil production units between 5 and 60 MW. This market is relatively small (about 3% of the Wholesale market) and almost constant in volume over the years (after 2003: 275 MW yearly contracted capacity for secondary control ‘regelvermogen’, and 300 MW for emergency reserve). The secondary reserve bought by TenneT on-line from the active market players is activated automatically (Automatic Generation Control). Two types of market players can be distinguished on the Balancing market, active and passive market players. Within the active market players, again two types can be distinguished, players with secondary reserve capacity contracted on an annual base and players without capacity contracted. According to the actual tender procedure of cross-border capacities (the last import/export energy is established one day-ahead), this will not have a direct impact on the price formation on this market. Other price formation drivers of the Balancing market are for example: unexpected unavailability of production units, unexpected changes in demand due to weather changes, unexpected energy flows due to plant failures in neighbouring countries and too much wind energy produced in the north of Germany. They influence the demand or supply bids and with that the price formation.
11
in the Netherlands known as “PV-partijen” Programme Responsible Parties have to send their E-program to TenneT 13 to be established by TenneT based on UCTE rules and norms 14 as much as necessary, most of the time above the annual contracted capacity 12
-23-
50680533-Consulting 07-0048
The passive players on the Balancing market, producers or PRPs can and do influence the short / long position of the system by making use of the on-line signal TenneT publishes on its website. They instruct controlled production units to produce more or less electricity.
2.3.2
APX15 (Amsterdam Power Exchange)
The APX is a day-ahead physical voluntary market, anonymous, transparent in volume and price formation (bid and ask curves are public on the APX website) with clearing and settlement facilities. This is the market place where parties can balance their portfolios. It started operation end of 1999. The volume of the spot (APX) market is relatively low (16.0 TWh in 2005); it would even have been lower if the imported capacity of the daily auction would not have to be (mandatory) traded on the APX. Each morning before 10:00 a.m. participants submit, via an electronic trading system, their purchase or sell bids for each hour of the following day. Each of the participants determines its bid-values based on market information available at that moment, like: •
available production units, wind forecast
•
selected long or short position of market players (as far as this is known)
•
import-export capacity and prices
•
weather measurement data, forecasting and influences, etc.
Based on market data and its own portfolio position, participants adjust their strategy by using their knowledge on each decision-making moment. The saying ‘learning by doing’ is very applicable to this market. The longer the trading experiences on a market place the larger the market knowledge for that specific market because each market place has its own particularities. The quality of the price forecast a trader makes depends on market information and his/her ability to deal with that information and to learn from it complemented with the size of the company he/she is working for and the feeling he/she has built up on the existence and division of the market power. Also the information a trader uses, for example simulation models and forecast software, provides more input to a solid prediction of the price behaviour. At market closure, all purchase and sales bids are aggregated by hourly periods and ranked by price. For each hour, the intersection of the aggregated supply and the aggregated demand curve determines the market results, i.e. the Market Clearing Price and the Market 15
Also known as day-ahead market or spot market
-24-
50680533-Consulting 07-0048
Clearing Volume. After this ‘fixing’ is performed, the market results are made available to the participants, including all the purchase and sell bids for each hour. This information again is used by traders to understand and learn from market behaviour. Hourly market prices and volumes are used as a basis for calculation of indices that are published on the website and can be used as reference prices for benchmarking purposes. Cross-border capacity has a direct effect on the APX prices as parties are obliged to sell all the energy imported via daily auction on the APX. Based on the latest information (2005) this volume is about 5 TWh which is some 30% of the APX 2005 volume, mandatory traded on the APX. The APX started an intra-day market as of 14th September 2006.
2.3.3
Over The Counter market
The OTC (Over the Counter) market is represented by several market places. International energy trading counter parties active on the Dutch Wholesale market are for example GFI group, Spectron and ICAP Energy AS. The Energy Future Exchange ENDEX became the most important market place within the OTC market in 2005 and will be presented in the following sub-section. On the OTC market standard products are traded, mostly in parts of 5 MW. Periods can vary from intra-day hourly trade to days, months, up to three years ahead. The intra-day trades are hardly transparent and prices are mainly driven by availability or free (over-)capacity of power plants. This results in limited market volumes. Market parties use bilateral long-term contracts in order to cover (sell or buy) the main part of their portfolio and as such try to reduce their risks. Standard products (base, peak, off-peak, weekend hours) are predominantly used, and sometimes profile contracts.
2.3.4
ENDEX Futures Exchange
The ENDEX Futures Exchange16 (ENDEX) is a recognised exchange and started operation in 2003. ENDEX offers in The Netherlands mainly power futures for standard products, OTC
16
considered as OTC market place
-25-
50680533-Consulting 07-0048
clearing services and supports the energy market with an electronic trading platform. The ENDEX is supported by four liquidity providers: Essent, Electrabel, Nuon and RWE. The ENDEX liquidity providers guarantee a constant supply of bid and ask prices within a certain spectrum. ENDEX creates market transparency and enables market participants to manage price, volume and settlement risk on medium (months) and long term (years).
2.3.5
Market regulation
In this paragraph we discuss the influence of market regulation and market performance on the value of wind power. Market regulation should ensure that the imbalance settlement rules for producers and consumers lead to imbalance costs being equal to realized system regulation costs, such that wind power producers are not penalized more than the actual costs incurred by the wind power prediction errors. The Dutch imbalance market is regulated by a two-price model, which makes a distinction between passive and active contribution. Active contribution to system balance takes place when Reserve and Regulation Suppliers provide balancing power through bidding on the imbalance market. Passive contribution takes place when a PRP produces (consumes) more or less than scheduled, and whereby the resulting imbalance is opposite to the system balance direction. In the current Dutch regulatory system, it seems that wind power producers (wind farm owners) are treaded as if they passively contribute to balancing the system and therefore are only paid the spot price for their electricity minus a fee for each MWh of wind power to pay for the “expected” imbalance of the wind farm. However, wind farms are part of the production portfolio of a PRP and might both contribute to the imbalance or reduce the imbalance of the PRP. The latter can be the case when the overall position of the PRP is long when the system is also long and, at the same time, the actual wind production is less then the forecasted wind production. The Dutch TSO charges or pays PRPs for their overall imbalance energy, without specifying the imbalance costs to individual producers, such as wind power producers. Only when a wind farm owner gets PRP status, the imbalance costs can be specific and will reflect the actual market value. However, current market regulation acts as a barrier for wind farm owners to be acknowledged as a PRP. As a result, wind farm owners are hand over to a
-26-
50680533-Consulting 07-0048
suboptimal situation, where the large PRPs determine wind power’s imbalance cost based on a method which is not transparent.
Table 2.2: Overview of the Wholesale prices on two market places TenneT Balancing Market (prices in Euro/MWh) Year
upward (RRV) 17
downward (RRV)
PV short18
APX (prices in Euro/MWh) PV long
2000
day-ahead 48
2001
71
2
-
-
33
2002
63
5
34
19
30
2003
100
3
55
35
46
2004
71
8
44
30
32
2005
91
13
53
38
52
Gate closure time of the day-ahead market is another important aspect of market regulation. Some electricity systems allow an intraday market, which enables wind power producers to trade against their latest and most accurate production forecasts. Until recently the Dutch TSO did not operate an intraday market, which forced wind power producers to trade on forecasts with a longer time horizon (this changed in September 2006). Even when intraday trading is facilitated, this does not necessarily mean that more accurate wind forecasts lead to lower imbalance costs. When an intraday market is not liquid enough, a wind power producer cannot find a counterparty to sell its energy. Studies from DTe (2005a; 2005b), the Newberry et al. (2003), OSCOGEN (2001) have been used to assess the performance of the Dutch wholesale market on liquidity, transparency and degree of competition. The general conclusion of these studies is that the Dutch wholesale market scores average on liquidity compared to other European countries. A strong improvement is possible and vital for the functioning of the Dutch electricity market. There is still a lack of publicly available demand information, actual usage and nomination of interconnection capacity, day-ahead nominations, ex-ante and ex-post production information and trade results of the OTC market. Furthermore, the composition of 17 18
generation units qualified to provide upwords regulation (RRV = regel- en reserve vermogen) means that the system balance (sum of imbalance volumes of all PRPs) is short
-27-
50680533-Consulting 07-0048
participants has changed in the last few years. The number of pure traders (without a physical position) has reduced from 12 in 2002 to 4 in 2004. The number of vertical integrated participants has remained constant. It can be concluded that the concentration levels on the Dutch wholesale market is very high, which might indicate that presence of market power is plausible.
2.4
Wind power characteristics
Relation between wind speed and wind production Wind turbines extract kinetic energy from moving air, and convert it to mechanical energy in the turbine rotor, and further to electrical energy through the generator. The kinetic energy of the wind, flowing through the turbine rotor (propeller), is per unit of time the power Pwind: Pwind = ½
airSrotorvwind
3
[Formula 2.1]
the mass density of air; Srotor the propeller area; vwind the wind speed. air
Formula 2.1 shows a third order relation between wind speed and wind production. This implies that large upward variations in wind power production are more frequent than large downward variations. If we do not consider forecast quality, wind power will need more rampdown power than ramp-up power. This is an important factor on wind power’s imbalance costs as ramp-down prices are generally lower than ramp-up power. Type of wind turbine The type of wind turbine also influences wind power’s imbalance costs. The wind turbine technology determines the power curve, which is function of wind speed at hub height (i.e. rotation centre of the rotor). The power curves of the two most common wind turbines are shown in figure 2.1.
-28-
Figure 2.1
50680533-Consulting 07-0048
Typical power curve of a fixed-speed stall controlled turbine (a), and a variable speed pitch-controlled turbine (b).
The variable speed pitch controlled turbine shows a constant wind production at wind speeds higher than 16 m/s. This is a great advantage over the fixed-speed stall controlled turbine, because the controllability of power production increases. More controllability of power production will give wind power more value. Most new types of wind turbines have improved power curves for wind speeds above the cutout wind speed (approximately 24 m/s). Instead of a vertical line representing a sudden drop to a production of zero (like in figure 2.1), these modern wind turbines gradually reduce their power level in a time period of 5 to 10 minutes19. Wind turbines with this power curve will lead to lower imbalance costs, as the power curve demands less ramp speed of the conventional capacity. In this study, we use the power curve of the variable speed pitch controlled turbine as basic input for our analysis quantifying the value of wind power. Variable power output An important characteristic of a wind power plant is the stochastic power production. The variability in power output emerges through fluctuations in wind speed. A wind power plant can therefore be seen as non-firm capacity, which can jeopardize system stability. The impact of wind power’s variability on system stability depends on the time-scale in which the variations occur.
19
time period and cut-out wind speed can be made different per wind turbine in a wind farm
-29-
50680533-Consulting 07-0048
In a minute-to-minute time-scale the variations in power output will influence the primary control mechanism. A higher frequency and volume of short-term power output fluctuations could lead to a shortage of balancing power. In such a situation, conventional power plants responsible for providing primary control will have to increase their balancing power by operating at suboptimal production levels. When this situation occurs the fuel, maintenance and imbalance costs will increase (KEMA, 2005). However, many studies conclude that wind power will not effect short-term system balancing, either because they are small compared to other disturbances of the power balance (load variations), or because they even out over the large number turbines and therefore vanish (Kling & Slootweg, unpublished UWIG, 2003). Furthermore, minute-to-minute variations in the wind are smoothed out by variable speed technology, inertia of the large rotors and the variations in power output of the individual wind turbines within a wind farm. Therefore, the effect of wind power on system operation in the primary control time-scale is small even at considerable penetration (Ernst, 1999; Kirby et al. 2003). The extreme step changes recorded from one 103 MW wind farm are 4-7% of capacity in a second, 10-14% of capacity in a minute and 50-60% of capacity in an hour (Parson et al., 2001). The maximum measured change in output from 2,400 MW of wind capacity in western Denmark is about 6 MW per minute (Christensen, 2003), which is only 0,25%. The difference between the measurements of the 100 MW wind park and the 2,400 MW wind portfolio shows that aggregation of wind farms leads to a reduction of wind power’s variability. When we consider wind power’s variability on an hourly time-scale, the wind power plant will affect the secondary control mechanism. Wind power has an effect on the total amount of load-following reserve capacity if the maximum of net load variations (wind included) is larger than the maximum of load variations (wind excluded).
-30-
50680533-Consulting 07-0048
Hourly variations of offshore wind power (100 MW)
Number of hours
10000
1000
100
10
1 -100-90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 100 Percentage of w ind capacity
Figure 2.2
Wind power’s variability
Figure 2.2 shows that the hourly variations of a 100 MW offshore wind farm are within -25% to 25% of the total installed wind capacity during 99% of the time. These findings correspond with a study of Milborrow (2001) which concludes that the maximum hourly change in output power from distributed wind rarely exceeds 20% of the installed capacity of the wind farm. Furthermore, large upward variations are more frequent than large downward variations. This is the result of the third order relation between wind speed and power output (P ~ v3). In The Netherlands, the net hourly load variations of the total system are between -1,200 and 2,700 MW during 99% of the time. If we assume that the hourly wind power variations remain -25% to 25% of total wind capacity by higher wind penetration levels, we can calculate the maximum wind capacity that could be installed without creating variations larger than -1,200 MW and 2,700 MW. This method results in a total wind capacity of 4,800 MW, which varies for 99% of the time between 3,600 MW and 6,000 MW. This is at the pessimistic side, because the hourly wind power variations will reduce when more capacity is installed due to the aggregation effect of multiple wind farms. However, the maximum of net hourly load variations will increase when a wind portfolio of this magnitude will be integrated. In the worst case scenario, the wind capacity simultaneously varies with the system load in the opposite direction, e.g., wind capacity varies with -1,200 MW, while the system load varies with 2,700 MW
-31-
50680533-Consulting 07-0048
Unpredictable production One of the most important characteristics of wind power is its unpredictability. The primary energy source of wind power, the wind resources, is hard to predict. Forecasting the electricity production from wind power is possible only to a limited extent and the forecasting quality strongly depends on prediction window. In the current Dutch regulation scheme, wind power producers are practically forced to predict wind power 12-36 hours in advance, as there is practically no intraday market. This prediction window results in relatively high forecast-errors. However, from September 2006 the APX will started an intraday market, which can significantly improve the quality of wind power prediction, resulting in smaller forecast errors. Even if the average quality of the forecast is very good, there will be times at which the forecast error deviates strongly from the average. In a position paper of the UCTE on wind power it is argued that “not the yearly average but the individual maximum forecast deviation determines the value of power/energy reserves to be programmed by the transmission system operators in order maintain system stability”. Figure 2.3 shows the frequency distribution of the hourly forecast error for the day-ahead forecast of a simulated 100 MW offshore wind farm.
Hourly absolute offshore forecast error distribution (100 M W)
Number of hours
10000 1000 100 10 1 1
11
21
31
41
51
61
71
81
91
Percentage of capacity
Figure 2.3
Frequency distribution of the hourly forecast error for the day-ahead forecast of a simulated 100 MW offshore wind farm
-32-
50680533-Consulting 07-0048
Forecast error distribution The figure shows that 90% of the forecast error is within a range of 20 % of total capacity. Figure X also shows that there is one hour during the year where forecast is off by some 70% of the total installed wind capacity. Important to note is that the impact of a forecast-error can be decomposed in a volume and time component. The deviation between scheduled and actual production represents the volume of imbalance energy that needs to be compensated. The time and the speed at which the deviation between scheduled and actual wind production takes place determines whether the conventional generating units are able to follow the fluctuations in wind speed production. During night hours a large forecast error results in a higher risk for system stability than when the same forecast error occurs at day-time during peak-load. Availability Wind power’s availability significantly differs from thermal and hydro generation, because the primary energy source cannot be stored and is uncontrollable (Kling & Slootweg). The availability of wind power includes all planned and unforeseen outages, which are the result of maintenance, turbine failures or extreme weather conditions. Although the technical availability of onshore wind turbine is relatively high, around 98%, the uncertain availability of wind resources significantly reduces the overall availability. The availability is an important characteristic of power plants because it determines the long term balancing capabilities of the system. Wind power has a negative influence on the long term system reliability, because long-term predictions of wind power production are unreliable and even impossible. Extreme weather conditions such as heat waves or storm fronts, which lower wind power’s availability to zero, cannot be predicted upfront on the long term. Consequently, the long term balancing of the system becomes riskier.
2.5
Valuing power plants
The value of a power plant can be determined by its ability to generate power, its generation costs, its influence on the electricity wholesale price and its contribution to system reliability. The value of power generation is predominantly dependent on fuel prices and therefore most studies use avoided fuel costs as derivative for the value of wind power. The amount of avoided fuel can be determined by the capacity of conventional power plants, which can be displaced by wind turbines.
-33-
50680533-Consulting 07-0048
Apart from their value as energy producing entities, power plants represent a capacity value. The capacity value of a power plant can be defined as the power plant’s contribution to system reliability. For every additional MW of generation capacity that is added to the system the risk on a loss of load situation reduces. However, the capacity of power plants differs in quality. Therefore, it is important to make a distinction between firm and non-firm capacity. Power plants with firm capacity can control their production level and have a high availability. Power plants with non-firm capacity, such as wind power, have limited control over their production level and an uncertain availability. The characteristics of non-firm capacity limits the ability of these power plants to contribute to system reliability, instead they can even reduce system reliability in some occasions. In this study, we will give an overview of the arguments and methods that attempt to value wind power on its energy and capacity value. Concerning the active power of wind power in a power system, one quantity that indicates the value of a wind turbine or aggregated wind park is the capacity factor:
CF =
E yr Prated ⋅ 8760
in which: CF denotes the capacity factor Eyr annual energy production [MWh] Prated Rated Power [MW] This quantity gives information about the annual energy delivery, but gives little information about the capacity value of wind power, because it does not include the impact of wind power production on system reliability. If the annual wind power output is mainly delivered during peak-load times the value of wind power will be high. However, in the opposite situation, when wind production correlates with off-peak hours the value of wind power will be relatively low. Thus, the capacity factor cannot be used as an indicator for wind power’s availability nor to derive a capacity value directly. Capacity credit An indicator that includes the wind power’s impact on system stability is the capacity credit. The capacity credit of wind power is the amount of installed conventional power generation capacity that can be replaced by wind power generators, without an increase of the LOLP.
-34-
50680533-Consulting 07-0048
The loss of load probability (LOLP) is the probability that the system load exceeds the reliable generation capacity. Typical values for the LOLP for modern power systems are between 2,4 – 4 hours a year. This does not include power import from neighbouring countries. Intermittent renewable generators typically have low mechanical failure rates, but are not able to generate power when the resource is not available. Therefore, wind power generators have a larger influence on the LOLP than conventional generating capacity. The calculation method of the capacity credit is based on comparing the LOLP of a system without wind power capacity with a system where wind power is integrated. By keeping the LOLP or system reliability equal to the reference scenario, the effective load carrying capability (EELC) of wind power can be calculated. Finally, the EELC is divided by the total capacity of the wind power plant, which results in the capacity credit. A more comprehensive methodology of capacity credit is given in appendix II. The capacity credit will increase with increasing wind power capacity, closely related to the capacity factor, but not proportionally. The additional capacity credit will become smaller with every additional installed capacity of wind energy. At low wind penetration levels, wind power’s contribution to meeting system load will be relatively high, because wind production will be available in addition to other power plants amongst other during (some of the) peakload hours. However, when more wind capacity is added to the system (while the load remains the same) the additional wind power production will contribute less to system reliability, as most peak-load hours are already covered by the existing wind portfolio and conventional capacity. Moreover, the capacity credit of wind power cannot be treated as a complete substitute of firm capacity. Although the wind production is available at peak-load hours, its variability will cause balancing problems, because all firm capacity is already committed and cannot be used to balance wind power’s variability. At certain penetration levels, adding more wind capacity can actually increase the LOLP, because of wind power’s variability Just like the capacity factor, seasonal effects influence the capacity credit. In figure 2.4, the capacity credit is specified for the winter season (a) and summer season (b). The figure also shows that increasing wind power penetration results in relatively smaller capacity credit improvements.
-35-
Figure 2.4
50680533-Consulting 07-0048
Capacity credit of the aggregated wind park for all scenarios, winter (a) and Summer (b)
A study of ESBNG (2004) shows how the capacity credit of a wind portfolio reduces the need for reliable gas production capacity. Figure 2,5 shows the plant requirements, which are needed to meet a peak-load of 6500 MW. In this example, a power system with only gasfired plants is compared with a power system in which gas-fired plants and wind power are combined.
Figure 2.5
Illustrative plant requirements without/with 2500 MW of wind (ESBNG, 2004)
-36-
50680533-Consulting 07-0048
It may be noted that the “apparent” plant margin increases when wind power capacity is installed. The plant margin can be defined as the difference between peak demand and reliable production capacity. A peak demand of 6,500 MW (first bar) requires around 7,800 MW of thermal plant (2nd bar), assuming a plant margin of 20% is needed to meet the security standard. (The exact percentage is not important). If, say, 2,500 MW of wind is installed, with a capacity credit of 429 MW – using ESBNG’s figure – then only 7371 megawatts of thermal plant is needed. Balancing the system When the imbalance volume of a wind portfolio needs to be compensated, a PRP can decide to self provide balancing power with its own generation portfolio. This will require pro-active portfolio management, i.e., operating power plants at partial load with suboptimal efficiency. The costs of self provided balancing power depends on the technical properties of the conventional generation portfolio and the trading position of a PRP. The rate of change and the efficiency by different output profiles are decisive in the operation of the plant and consequently the commitment of the plant. For example a PRP will rather operate two plants on 50% of nominal capacity than 4 plants on 25% of nominal capacity, which limits the reserve and regulation available in the system. The short- term volume risk can result in critical situations for system security when wind forecast errors occur at night hours or at the flanks of peak-load hours. During these hours there is relatively little conventional energy operational, which limits the possibilities of PRP’s to self provide balancing power. When a PRP is not hedged against this risk and cannot self provide the power needed to compensate wind power’s imbalance volume, TenneT will procure reserve and regulation power in the imbalance market. The imbalance price will depend on the marginal cost bid ladder of reserve and regulation power suppliers (RRS). The prices on this bid ladder are determined by the marginal costs of production, which are higher than the wholesale prices to compensate for the additional start and stop cycles and less annual operation hours. When wind penetration increases the average imbalance volume during each PTU will also rise, which will result in a higher average price for reserve and regulation power.
-37-
3
50680533-Consulting 07-0048
WIND POWER ON THE WHOLESALE MARKET
The value of large scale wind energy in a liberalised market is to a large extent determined by the price level on the wholesale market. Different elements influence this price level and are therefore important factors in determining the value of wind energy. However, in case large amounts of (subsidized) wind energy come on the market, this will affect the price level on the wholesale market. Most directly the effects will be visible on the spot market and imbalance markets, but it will also trickle through to future/forward prices and electricity options, as wind energy will change the volatility in the market and hence will introduce price and volume risks. In this chapter we explore the factors that influence the wholesale market price on the short and long term in general. Then we focus on the influence of large scale wind energy on the price levels of the APX-spot price and the imbalance price.
3.1
Factors that influence the electricity wholesale price
In a liberalised market, there are several electricity price formation elements and factors that influence the electricity wholesale price in a region or a country. For every moment in time electricity supply must meet the demand as there is no possibility to store electricity in The Netherlands at present. The most relevant factors are given below. The size and composition of the power generation park. Depending on the availability and strategy of power plants owners, the supply curve is determined (price and volume available on the market). The merit order of supply shows the price at which plants deliver electricity into the market. As it takes long to build new power plants, this can be taken as given in most cases. However newly introduced power plants20 will change the merit order. In general new power plants are only built if their long run marginal costs are lower than the long term Wholesale market price. The demand over time. Demand has a strong influence on the market price. Higher demand will be served by more expensive generation units. Demand varies in time. Besides a seasonal pattern, strong 20
including wind capacity
-38-
50680533-Consulting 07-0048
weekly and daily patterns can be recognised, resulting in different price levels as base load price, peak load price and off-peak load price of electricity. Further it should be noted that as a result of varying demand the market characteristics (price elasticity, price level and volatility) vary in time. Fuel prices. Fuel prices are important factors in the marginal costs of fossil power plants and therefore in the spot market price for electricity. We mention: – Crude and brand oil markets – Coal markets – Gas markets. In the Netherlands gas plants are the most expensive ones (both commodity and capacity fee) setting most of the APX price. This is partly caused by a strong link between gas and oil prices. It is expected that this link will become weaker in the future. The coal and oil market are effectively decoupled. The CO2 emission credits are becoming an important factor for the marginal costs of a power plant. The costs for CO2 allowances are closely linked with the type and quantity of fuel that is being used at a fossil power plant. Therefore, the CO2 costs can be seen as additive to the fossil fuel price. As a result the marginal costs of fossil power plants will increase, thereby changing the price settlement on the wholesale market. A smaller number of CO2 emission credits will result in higher prices for fossil fuel thus making wind power more competitive. “Cross border” capacity and prices. The spot market price in the Netherlands is influenced by the price formation in Germany, Belgium and France. A price difference can occur when the cross border capacity is limited. Annual, monthly and daily cross border capacity is auctioned by TenneT. In the future it is expected that more cross-border capacity will come available. When interconnection capacity becomes available between the Netherlands, UK and Norway, price formation in those countries or regions will influence the spot market price in the Netherlands. Reserve capacity margin. The reserve capacity margin is the margin between maximum load and peak generation capacity. When the reserve capacity margin is low, imbalance prices are high and LOLP increase The increase of installed wind capacity will have the following effects: - during periods with wind speed higher than some 4 m/s and wind capacity is technically available, the reserve capacity of the system increases;
-39-
-
50680533-Consulting 07-0048
during periods with wind speed lower than 4 m/s or higher than say 25 m/s, or when wind capacity is not technically available, the need for reserve capacity in the system is higher (forecasted and not produced wind capacity needs to be compensated).
Number of market players and market structure. The number of players active in the market, the market structure (regulations, etc.) and the liquidity in the market are important factors that influence the competition in the market. In case of thin competition, some players may exercise their market power to maximise their profits and keep the market price at a higher level. In the Netherlands DTE carries out market surveys to investigate the competitiveness of the market. For the short-term (minutes up to hour level) the following influence factors are the most important (deterministic) ones: – the availability of the power plants – the demand level – the wind forecast error (differences between wind speed forecast and actual wind speed) – the availability of the cross-border capacity – the demand forecast error (difference between demand forecast and demand actual).
3.2
The impact of wind power on the wholesale market
The integration of offshore wind power will introduce additional volume and price risk in the Dutch wholesale market, which will effect both short and long-term energy trading.
3.2.1
Effect on the APX-spot market
Large scale integration of (offshore) wind power will mainly affect the APX-spot market. Trading on longer time scales is risky considering the fact that wind forecast quality is strongly deteriorating for prediction windows longer than 36 hours ahead. Once wind farms have been installed, the marginal costs of wind energy are very low (no fuel has to be burnt). Therefore, based on wind energy forecasts it can be bid into the APX-market at a very low (zero) price. Since the market will settle at a higher price, wind power will be automatically be sold and receive the market price. Wind power acts as a “price taker”. On average one would expect that this effect results in a lower average market price, since “expensive” units are pushed out of the merit order.
-40-
50680533-Consulting 07-0048
However, since bids on the APX have to be entered 24-36 hour in advance the market will respond to the expected wind energy production during the next day. When wind power production is expected to be high the market will respond and bid at lower prices. However, when less wind is expected, traders will most probably respond in bidding higher prices, seeking to cover their capital costs during these hours. The consequence is that the market price volatility is expected to increase. The risk of large price differences will be especially high when supply meets demand at the steep part of the market bid ladder curve, i.e., moving from base to peak-load. When supply meets demand on a flat part of the market bid ladder curve (for instance during night hours) the price differences will be marginal. This situation is illustrated by figure 3.1, which shows that when demand meets supply at a flat part of the bid ladder curve the price will only reduce by P1, while at the steep part of the bid ladder curve the price will reduce by P2.
Demand
Bid
Bid curve
curve
curve
with wind
P (Euro/ MWh)
P1
P2 Import
Base-load
Peak-load V (MWh)
Figure 3.1 The results of the correlation analyses can be found in Appendix 2.
3.2.2
Influence on future markets (ENDEX)
Due to the unpredictability of the wind speed electricity generated by wind farms is mainly traded on the spot market. However, the integration of a considerable amount of wind energy will eventually decrease the average demand for fossil electricity, which could lead to lower
-41-
50680533-Consulting 07-0048
ENDEX prices. At the other hand, long-term portfolio hedging will become riskier as wind power’s long-term availability is uncertain. PRPs will respond on this additional volume risk by changing their long-term trading strategies. Risk adverse portfolio managers might schedule a larger part of the expected load on a short-term basis, which reduces the longterm stability and the transparency of the wholesale market. This would result in less baseload trade on the long-term markets, because portfolio managers will have to consider the base-load capacity delivered by wind power. This situation is illustrated by figure 3.2, in which the capacity credit of the national wind portfolio is part of the long-term base-load volume. A risk bandwidth illustrates the uncertain availability of wind power’s capacity credit.
V (MWh) Expected load Volume Risk bandwidth
Long-term Base-load Volume Capacity credit wind
T (months) Figure 3.2
The long-term volume risk of wind power can be hedged by futures and peak-performance capacity, in order to avoid imbalance costs or a suboptimal energy schedule. The price risks can be hedged by options and spark-spread contracts.
3.2.3
Influence on the imbalance market
Wind power will affect the amount and frequency of imbalance, as forecast-errors result in deviations between scheduled and actual energy production. As a result, the need for realtime balancing will increase, thereby driving up the price of regulation and reserve power, which affects the imbalance costs of all PRPs. When we account for the fact that wind generation can change significantly on a PTU time-scale, the volatility of imbalance prices over the PTUs might increase. It is at present unclear how frequent these sudden changes in wind generation occur. Literature seems to indicate that on average wind generation is rather
-42-
50680533-Consulting 07-0048
persistent and that there will be limited influence on the volatility of the imbalance market. However, dependent on the need for fast regulating power, it might be necessary to increase the requirements for the depth of the imbalance bid ladder. PRPs can also decide to increase regulating power by operating certain production units at partial load. This will enable them to self-provide regulation power when needed, or to participate on the imbalance market by offering balancing power. And these will increase the costs, so the market price.
3.2.4
Effect on fuel prices
Oil and gas prices are very sensitive to changes in demand and production. Indirectly, large scale wind generation may influence the market price of electricity through this mechanism. In case wind energy displaces large amounts of fossil power, this will lower demand for fossil fuels, hence resulting in lower fuel prices and therefore lower electricity prices.
3.3
Correlation between wind power and electricity prices
We statistically analysed one year of data from a former study21 carried out by TenneT, ECN, TUD, Ecofys and KEMA by order of Dutch utilities and sponsored by We@Sea. The data comprises wind speeds, APX prices and imbalance prices and volumes during the period of 1 June 2004 – 31 May 2005 with a resolution of 15 minutes. For the wind speeds we used wind forecast data from ECN and wind realisation data from weather stations of KNMI. The market information like imbalance prices and APX prices have been extracted from the APX and TenneT databases. We applied the data on a simulated 100 MW on-shore wind farm (on three locations, i.e. a national “average location”, De Kooy and Leeuwarden); the latter two being meteorological stations of KNMI. In our analyses we used correlation techniques to find the relationship between wind speed forecasts and the APX price and wind speed forecast errors (actual minus forecast values) and the system imbalance. We expect some statistical noise because of abnormal events which influenced the wholesale market in 2005. A detailed description of our analyses can be found in Appendix II. The main results are given below.
21
KEMA, 2006. Balanshandhaving met 6.000 MW aan windenergie
-43-
50680533-Consulting 07-0048
Only a few per cent of the variation in the APX price can be explained by variations in wind speed forecasts. However, in January and February 2005 between hour 0 and 6 a.m. on non-working days the hourly correlations of the APX prices and the on-shore wind speed forecast show relatively high negative values (-0,86 < R < -0,45). In one of the data series we also found high negative correlations during the same nightly hours in May 2005. During these specific hours occasionally 20-75% of the changes in the APX price is caused by wind power. This can be partly explained by the coexistence of relatively high wind speeds and low APX prices during nightly hours. Furthermore, during these hours the system switched over from “night load” to “day load” which means that the APX price was settled at the steep part of the market merit order. Thereby wind power production could lead to relatively large price reductions. To conclude, only during very few hours per year wind speed forecasts – indicating the wind power production – has a significant price reducing effect on the APX. During all other hours this effect is hardly recognizable. The correlation analyses have also been used to explore the relationship between wind speed forecast error and system imbalance. The first conclusion on these analyses is that a significant positive correlation is shown occasionally between short system balance – when electricity consumption is higher than the electricity consumption Programme Responsibility Parties have been forecasted – and a negative wind speed forecast error (more wind power has been produced than forecasted). The frequency of occurrence of values R2 > 0.2 is about 3.3% of the time. Significant positive correlations also is shown occasionally between long system balance – when electricity consumption is lower than the electricity consumption Programme Responsibility Parties been have forecasted – and the positive wind speed forecast error (more wind power was produced than forecasted). For these situations the frequency of occurrence of values R2 > 0.2 is about 3.8% of the time. Most of the abovementioned correlations have been found in sub series of nightly hours during non-working days, i.e. during hours with relatively low electricity demand. The overall conclusion is that, in general, the monthly and yearly correlations show that the wind power portfolio installed in 2004-2005 has almost no effect on the APX price and little effect on the TenneT imbalance market.
-44-
4
50680533-Consulting 07-0048
THE VALUE OF WIND POWER
Since wind turbines have been introduced in the electricity system, there has been debate on the value of wind power. In many studies the value of wind power was attributed to avoided costs, such as avoided fuel costs and/or avoided capital cost for investments in additional fossil plants. It was recognized that in order to keep the system reliable “spinning reserve” or plant margin should be maintained, of which the costs should be allocated to the wind power production units. These studies were most of the time using a marginal generation cost approach to determine the value of wind power. Given the high capital costs of wind projects, financing wind farms was only possible using fixed power purchase agreements over (a large part of) the lifetime of the wind turbine. The liberalization of the electricity (wholesale) market at the turn of the century changed the way wind power was valued. No longer could wind power be valued using SEPs centralized production units and reserve power, but PRPs had to find the value of wind power within their portfolio. Although wind power was virtually sold in the APX-power market, and had to be balanced on TenneT’s imbalance market, most PPAs still used fixed contract prices over a long (10 years period). Traders became aware of the risks involved in having wind power in their portfolio. More recent, following strongly rising prices on the APX, wind farm developers are willing to take the risks of fluctuating revenues, because they want to profit from the upward potential in the electricity price. Although banks are still hesitant to follow this approach, there are signs that more parties will take make this move. Financial instruments could be used to hedge the downward risk of wind farm projects. TenneT has put effort in making the imbalance market more liquid, trying to lower the price of regulating power. Some PRPs are moving towards full transparency, offering wind farms (virtual) access to the imbalance market. As a result imbalance costs are moving closer to the APX prices. PRPs state that the imbalance market has become so transparent that they offer all reserve power to TenneT, instead of withholding power for their own balancing, as this shows to be a more effective mechanism and financially more efficient. If this is indeed the case, then the Netherlands have succeeded in developing a joint reserve power pool. When the electricity market becomes more mature, and electricity derivatives become more common it will be easier to derive the wind power value and the cost of related risks. However, these instruments such as power futures, put and call options are still developing.
-45-
50680533-Consulting 07-0048
It will therefore take time before these instruments can be used to make a full assessment of the components that constitute the value of wind. In this report we will therefore stick to the more common approach and argue that a fundamental model can be used based on marginal cost of power generation and demand.
4.1
Overview of value and cost components
This paragraph gives an overview of the value and cost components of the wind power price. For each price component, we explain the key drivers and describe a methodology for quantification. Where possible we have used historical market prices and wind data to make the first rough estimate of the price component’s value. In order to quantify the components we have used the data series already described in section 3.3. For the quantitative analyse in this section we also used the aforementioned case studies of the 100 MW on-shore wind farms and, besides, a 100 MW offshore wind farm. Furthermore we have used a short-term perspective, which means that we assumed that all wind power produced is sold on the APX. Currently, this reflects the real market situation, but it is not unlikely that wind power will also be sold on the ENDEX and OTC markets in the near future
-46-
50680533-Consulting 07-0048
Table 4.1 shows all value and cost components of the wind power price. Each price component is influenced by one or more factors presented in the first row of the table. The marks in the cells indicate a potential relation between price component and factor. The first two components is this section – market value and MEP subsidy – contribute to the price of wind power in a positive way. They easily could be regarded as value components, but one could get the APX price for electricity produced by any source. The market value thus does not distinguish wind from other sources. The MEP is an important price component of wind power closely (inversely) related to the market value, i.e. the higher the latter the lower the MEP. The MEP therefore is an output parameter rather than an input parameter. More details on these two components have been described below. Table 4.1: Value and cost components matrix Price components
Market value MEP (subsidy) Value components Image PRP (more customers) Wind pattern value Geographical spread Fuel diversification Plant Margin Effect (capacity value) Down regulating power Cost components Short term imbalance cost Long term risks Administrative costs
Wholesale market
Quality/ Accuracy of forecast
x x
Variability
x
Availability
Wind penetration
x
x x
Fossil production mix
System balance
x
Load
x
oil/ gas prices
Customers preferences
Legislation and regulation
x x x
x
x x x
x
x x
x
x
x x
x x x
x
x
x
x
x
x
x
x
x
x
x x
x
x x
x x
x
x
x
x
x
x
x x x
-47-
4.2
50680533-Consulting 07-0048
Value components
Image PRP (Vimage): by contracting wind power or integrating wind power in their portfolio a PRP can improve its image. A greener image may result in more clients (end consumers)) or in a premium price for green electricity. However, the value of a larger number of clients as a result of a greener image has in practice never been allocated to wind power, because it is very hard to demonstrate or quantify the value of this component. A factor that influences the value of this component is customer preferences, as the customer can compare the percentage of green electricity generation between various energy providers. In the Netherlands, the market for green electricity was liberalized first, which showed that a substantial amount of households switch suppliers. However, it is unclear whether this was caused by the “greenness” of electricity, or because the preferred a new supplier over the traditional one. However, in this stage of our study we do not quantify the value of this component, because the weight of this component on the total value of wind power is expected to be very small. Fuel diversification (Vfuel): adding wind power to a PRP’s production portfolio can reduce the risk for gas price volatility. When the wind power production pattern is correlated with price spikes in the gas market, a PRP can use its wind power resources to hedge against price risk in the oil and gas markets. In this way a wind power plant can make a positive contribution to the portfolio risk of a fossil generation mix (see also Awerbuch & Berger, 2003). In order to quantify the fuel diversification value of wind power, we need to focus on the longterm fuel hedging activities. In general, price volatility on the gas market can be hedged by two methods. The first one hedges price volatility by trading on spark spread22. This allows PRPs to sell electricity when variable costs (mostly fuel price) plus spark spread, expressed in euro/MWh, are lower than the electricity price and sell fuel when the variable costs plus spark spread are higher than the electricity price. The value of the spark spread premium is based on long-term speculations about fuel and electricity prices. Another method to hedge or mitigate the long-term price risk is by trading in options. Options allow a fixed rate for buying or selling a certain energy volume and are valid for a fixed duration.
22
The differential between the price of electricity and the variable cost of a production unit which produces electricity from natural gas, expressed in Euro/ MWh
-48-
50680533-Consulting 07-0048
In order to quantify the fuel diversification value of wind power we will use the value of spark spread and options, which are traded on long-term markets. The price of these market products includes the estimated (speculated) risk premium for price volatility in oil and gas prices. When we multiply this risk premium (euro/ MWh) with the annual wind production during times when the gas prices were higher than the electricity prices and divide this amount by the total annual wind production we get the fuel diversification value of wind power, expressed in Euro/ MWh. However, in the scope of this report we will not yet quantify this component. Geographical spread (Vspread): when wind power is offered from wind park farms on different locations, the fluctuations in the portfolio will be dampened because different wind speeds at different locations will partly reduce wind power variations. When at one location the wind speed decreases, the wind speed at another location can rise. Furthermore, there is a time lag between different locations, i.e., it takes a low pressure area several hours to pass all wind parks in the Netherlands. Therefore, the aggregation of wind speed patterns of multiple wind farms at different locations will smooth prediction-errors (Pinson et al., 2004). Another positive effect of geographical spread would be that the aggregated power curve of multiple wind farms, shows a gradual reduction of wind production when cut out wind speeds are reached. This will give the secondary control reserves more time to ramp-up their power production. As a result, the conventional generation system can balance higher wind penetration levels. We can conclude that wind park owners of multiple wind farms should have a lower volume risk premium than owners of a single wind park. Factors that influence the value of this component are the degree of geographical spread and the wind penetration level. When the wind park locations are highly concentrated the effect of geographical spread will be much smaller than when the wind parks are distributed over a large area. When wind penetration increases the smoothing out effect becomes less, as the wind speeds of different locations become more correlated. In a KEMA study23 a first attempt was made to quantify this effect for offshore wind farms in the Netherlands. The study indicated that the value of geographical spreading was rather small with a value in the order of 1 EUR/MWh. Apparently the distances in the Netherlands are not large enough for this effect to be substantial.
23
Functional requirements of offshore wind farms, Cleijne et al. KEMA report, 1998.
-49-
50680533-Consulting 07-0048
Plant Margin Effect (Vcapacity): The plant margin effect refers to the maximum amount of reliable capacity, which is available at peak load. This capacity margin ensures system security when unexpected outages occur. When more capacity is added to the system, the capacity margin will increase (when load is kept the same) leading to a higher long-term system security. The contribution to system security of new capacity is expressed in the capacity value. When wind capacity is added to the system, the capacity margin of the system will increase by the effective load carrying capability of the newly added wind portfolio. The effective load carrying capability of a wind portfolio is a percentage of the total wind portfolio, which will be available at the same reliability as firm capacity24 and is referred to as capacity credit. The capacity credit is calculated by a probabilistic approach, which defines reliable capacity, as capacity which when added to a system does not change the loss of load probability25. The capacity credit will increase by higher penetration levels, but not proportionally as the relative contribution to system reliability becomes smaller for each additional MW of wind capacity. In other words, more wind capacity will increase the capacity margin, but the growth in capacity margin will lead to relative smaller improvements in the loss of load probability. Because the capacity credit is based on a probabilistic calculation method which defines reliable capacity using historical data, it can only give statistical estimates on the available capacity credit and its uncertainty interval. This makes it very hard to quantify the capacity value of wind power based on its capacity credit. Still, we can try to isolate the capacity value of wind power by modelling the power system with and without wind power, while keeping all other variables constant. This would allow us to compare the costs of long-term traded volumes in a scenario with and without wind capacity. The difference between these two outcomes divided by the total production of wind power give an indication of the capacity value expressed in Euro/MWh. The calculation method is further explained by the following procedure : 1. Model the long-term traded base-load volumes and long-term base-load prices for a scenario with and without wind capacity by using ProSym. 2. Compare the total cost of long-term traded base-load of both scenario and calculate the difference 3. Divide the difference by the wind production.
24
firm capacity can be defined as reliable capacity in the sense that it has a high availability the loss of load probability expresses annual number of hours at which production was not able to meet load. 25
-50-
50680533-Consulting 07-0048
Wind Pattern value (Vwindpattern): in general, the average wind power generation is higher in winter than in summer. Also, the wind power generation shows a daily periodical behaviour, most pronounced in summer, with the highest generation in the late afternoon and the lowest in the early morning. Both seasonal and daily periodicities more or less coincide with the normal load fluctuations in a power system, with the highest loads in winter and in the late afternoon (Soens, 2005). This would give reasons to believe that wind power capacity represents a certain load following value. The value of this component is based on the relation between the market value (APX price) and hourly wind production (wind pattern) forecast. Factors that influence this relation are wind power’s variability, availability, the spot price and the system balance. Down regulating power (Vdown regulation) Modern wind turbine technology and park operation increases the controllability of production. This offers a wind farm owner an option to provide balancing power to the imbalance market, by regulating down wind farm production. Hereby, a wind farm owner can maximize its revenues. However, the price paid for down regulation power must be rather high to compensate for the opportunity costs that result from less production hours. In the current Dutch regulation scheme, the down regulation price must be higher than the MEP subsidy, in order to commercially down regulate a wind farm. Factors that determine the value of this component are the correlation between real-time system imbalance and realtime wind production, the down regulation price and PRPs portfolio management. The wind must blow during times the system is long in order to have down regulation potential and a PRP must be willing to effectively trade wind power on the imbalance market. In order to quantify this component we have assumed the hypothetical situation where a PRP will maximize the revenues of a 100 MW offshore wind farm. Furthermore, we used the historical imbalance prices of the year 2004/2005 and assumed a break-even point for down regulation feasibility by a down regulation price which is equal to the MEP26. We also assumed that all energy surpluses in the system will be regulated down by the 100 MW offshore wind farm as long as the wind production is sufficient. In this hypothetical situation a wind farm owner would earn between 2,4 and 3,5 Euro/ MWh for its down regulation services over the period 2004/2005. When the MEP subsidy becomes abolished, the down regulation value will be higher, because there are no opportunity costs. In this case, a wind farm owner would regulate down 26
The current MEP subsidy for offshore and onshore wind power is 97 Euro/ MWh respectively 65 Euro/ MWh.
-51-
50680533-Consulting 07-0048
its power production when the down regulation price becomes negative, which results in a down regulation value between 4,4 and 5,5 Euro/ MWh over the period 2004/2005. For the interpretation of these results it is important to keep in mind that the hypothetical situation, which is used to calculate the down regulation value of wind power, is not very realistic. In reality the conventional production units will already be able to down regulate production at positive down regulation prices, because of their avoided fuel costs. Moreover, a PRP will not down regulate wind production and pay the wind farm owner, when it can maximize its own revenues by regulating down conventional production units. As long as PRPs are responsible for scheduling and trading wind power production, it will be very unlikely that a wind park owner will earn revenues on the imbalance market. Still, we consider it important to quantify the potential of this value component. Besides, from an environmental point of view, one should first down regulate fossil fuelled power plants and, thus, optimize the use of wind power and other renewable energy sources.
4.3
Cost components
The major integration costs of offshore wind power are imbalance cost. We define imbalance costs as volume and price risk of wind power, which will affect both short-term and long-term energy trading. We cannot separately quantify volume or price risk, because they are interrelated and therefore impossible to isolate. Large volume fluctuations will result in higher price volatility and large price fluctuations will have an effect on how energy volumes are traded. Therefore, we will only make a distinction between short-term and long-term imbalance cost, which comprises the effect of volume and price risk combined. Short-term imbalance cost (Cimbalance): The short-term imbalance costs of wind power mainly concern the unpredictability and variability in wind production, which have an effect on the imbalance volume and the imbalance prices. The imbalance volume of wind power is the result of a forecast error, which represents the deviation between actual and forecasted wind production. The imbalance price is dependent on many factors of which the system balance, imbalance direction and the price for reserve and regulation power are the most important. The balance direction of the system (short or long) determines whether wind power is causing imbalance or is passively contributing to maintaining system balance. When wind penetration increases, the magnitude of the forecast errors will increase. This will lead to the commitment of expensive reserve and regulation power, thereby driving up the imbalance prices. When we consider the combined effect of both high wind penetration and variable production, it becomes clear that the volatility of imbalance prices will increase. As a
-52-
50680533-Consulting 07-0048
result, the imbalance market will be riskier for all PRPs, because the chance on higher imbalance costs increases. Finally, the regulation framework influences the short-term imbalance costs. When the regulation framework supports the operation of an intraday market, PRPs can optimize their physical position according to the latest wind production forecasts. Whether the latest forecasts can also effectively be traded in order to reduce wind power’s forecast error, depends on the liquidity of the intraday market. When there are no participants or the traded energy volume is very small, the latest wind power forecasts will not result in lower imbalance costs. For more details see section 2.3.5. The regulation framework also determines the allocation of the costs resulting from a forecast error. In Germany all costs of wind production forecast errors are allocated to the end consumer, while in the Netherlands the costs are allocated to the PRP’s. To quantify this component we will only take into account the cost that result from a forecast error and ignore the cost that result from higher price volatility. However, in most cases, a PRP charges a fixed fee on all production hours of wind power. A fixed fee does not represent the actual market value of a wind forecast error. To determine the forecast error we have used the wind speed forecast data of ECN and the wind speed realization data of DEWI, TU Delft, KNMI and ECN. By applying a basic power formula for wind power, we calculated the power output (MW) per PTU and converted this value to MWh/ PTU. Important to keep in mind is that the power output can never be higher than the nominal output of a wind turbine and power output will fall to zero when the cut off wind speed (24 m/s) has been reached. Between wind speeds of 14 m/s and 24 m/s, the wind production will not change, because a wind turbine produces at a constant nominal level. As a result, an actual wind speed of 18 m/s will not lead to a change in power production when the forecasted wind speed was 16 m/s; both wind speeds end up with the same production level. Therefore, we set the forecast error on zero for wind speeds between 16 m/s and 22 m/s and kept a 2 m/s safety margin both ends of the nominal speed range. As the average wind speed forecast error is less than 2 m/s, this safety margin is sufficient. The corrected forecast error was multiplied by historical imbalances prices of the year 2004/2005, to find the imbalance costs. It is important to note that this calculation method does not take into account the opportunity costs of having a positive forecast error in a system which is long. In Appendix V a more detailed description of this calculation method is given.
-53-
50680533-Consulting 07-0048
A 100 MW offshore wind farm resulted in 2.2 Euro/ MWh imbalance costs, while the imbalance costs of a 100 MW onshore wind farm ranged between 5.6 and 8.8 Euro/MWh. For all data series, the onshore imbalance costs represented approximately 20% of total wind production income on the APX. For offshore wind power the imbalance costs were only 6% of total wind production income on the APX. The difference between the imbalance costs of onshore and offshore wind power can be explained by two factors. Firstly, the third order relation between wind speed and power production results in a higher production volume per operational hour for offshore wind farms, because the average wind speed is higher on sea than on land. Secondly, the standard deviation of offshore forecast errors is somewhat lower than the standard deviation of onshore forecast errors. It is likely that the wind blows more constantly (smoothly) above sea than above land. . The long term risks (Clong-term): when more wind capacity is added to the system, the trading on long-term electricity markets becomes riskier. As wind power is always sold when available, long-term traders must consider a certain amount of wind production as baseload. However, the availability of wind power long-term timescale is uncertain. The best indication long-term traders have is the capacity credit of the national installed wind portfolio. Consequently, the risk on wrongly scheduling energy volumes becomes higher. Wind power therefore reduces long-term stability and transparency of the system. Traders will respond on this long-term volume risk by scheduling more energy on a short-term basis, thereby reducing the annual traded volumes of base-load capacity. This effect will become stronger when wind penetration increases, as a larger part of the base-load will be provided by wind power. Although wind power will provide a large part of the base-load volume, it remains uncertain which part. Therefore, traders will hedge the bandwidth of base-load volume, which becomes much larger because of wind power’s uncertain availability and variability. Hedging will probably be done by buying or selling options. This long-term volume risk of wind power will also influence short and long-term pricing. The price of long-term base-load products will increase, because the volume traded will reduce. Moreover, the trade in base-load options will include a higher risk premium, to cover the long-term volume risk. Increasing long-term speculation about wind power’s availability and lower annual traded volumes will make the long-term base-load market less liquid, which also increases the prices. Parallel to the effects on long-term base load market, wind power will have an effect on the long-term market for peak performance. Higher wind penetration will increase the need for
-54-
50680533-Consulting 07-0048
short-term balancing power, which is provided by peak-performance capacity. Peak performance capacity is mostly traded on the long-term market, as prices rise sharply on a short-term basis. A higher traded volume of peak-performance capacity will increase both the short and long-term prices. To quantify this component we have to model the long-term energy trade and long-term energy prices for a system with and without wind capacity. The long-term trading strategies as described above can be modelled in Symbad. Then we can compare the value of long-term hedging products (options) between the reference scenario and the wind scenarios. Administrative cost (Cadmin): theoretically we could also distinguish the program handling costs, which could be allocated to wind power. In practise, these costs are included in the total fee expressed in a Power Purchase Agreement. According to some wind farm owners the administrative costs are not transparent. Handling an E-program involves costs, such as transaction costs, manpower and supporting facilities. A PRP could also ask a fee for forecast responsibility. The value of this component is strongly dependent on the regulatory framework, which prescribes the conditions of program handling. Furthermore, the number of transactions needed to schedule wind power in the most efficient way will influence the administrative costs for program handling.
4.4
Discussion and preliminary results
In this chapter we have defined value and cost components of wind power relative to the spot price of electricity. The list of defined price components is not exhaustive. The relative low environmental and social impact of wind power compared to fossil power plants could be seen as a value component. The environmental and social costs of fossil power plants have been acknowledged by policy makers, but regulation has not yet been created to allocate these costs. However, part of the environmental costs is incorporated in the wholesale price since power producers have to pay for CO2 emissions if they exceed their limits of CO2 emission credits. To value the price components we focused on the components with a lot of influence on the value of wind power, e.g. imbalance cost of wind power. Hereby, we mostly used standard spreadsheet and database software to make the calculations. Modelling in Prosym and Symbad was not included in this phase of the study, which prohibited the valuing of price components, like fuel diversification, capacity value and long term risks. Table 4.4 shows the volume weighted wind power price and the price components which have been quantified for three different data series. Data series A, is based on the We@Sea
-55-
50680533-Consulting 07-0048
(2006) study, which uses average inland and offshore wind speeds of The Netherlands. Data series B and C are specific wind farm locations, which are used for reducing possible bias in the outcomes. Table 4.4: Overview of the cost and value components analysis
100 MW wind capacity
Data series A
Data series B
Data series C
(National average wind
(Wind speed at wind
(Wind speed at wind
speed)
park ‘de Kooy’)
park ‘Leeuwarden’)
Onshore
Offshore
Onshore
Onshore
Volume weighted wind power price (Euro/ MWh)
38.9
35.6
34.4
36.5
Down regulation value (MEP included) (Euro/ MWh)
3.5
2.4
2.7
2.8
Down regulation value (MEP excluded) (Euro/ MWh)
5,5
4,9
4,4
4,5
Short-term imbalance costs (Euro/ MWh)
-8.8
-2.2
-5.6
-6.8
Imbalance costs as percentage of total revenues (%)
22
6
16
18
The first row of table 4.4 expresses the volume weighted wind power price. As already indicated in this chapter, we do not see the electricity price of wind power as a value component. However, analyzing the value of wind production on the APX market provides interesting information. Based on the wind production time series and market prices of 2004/2005, we have calculated the annual revenues of selling 100 MW of onshore and offshore wind power on the APX. It is important to note that the volume weighted market value is based on wind production forecasts, because the APX market clears at 12 to 36 hours before real-time. Data series A shows a volume weighted wind power price of almost 39 EUR/ MWh for onshore wind power, while the average market price is 36 EUR/ MWh. The difference of 3 EUR/ MWh can be explained by a relation between onshore wind production and high price hours. This relation is absent by offshore wind power. A detailed analysis of the data showed that onshore wind production was stronger related with the load pattern than offshore wind production. A possible explanation could be that heating of the earth increases wind speeds during day-time, when peak-load prices are settled. Following this explanation, the difference of 3 EUR/ MWh could be designated to wind pattern value. See also section 4.2. However, the difference of 3 EUR/ MWh could also be expressed in other value components. Moreover, the relation between wind production and load has not been found by onshore wind power data of series B and C.
-56-
50680533-Consulting 07-0048
The second and third row of table 4.4, show the down-regulation value of wind power, with or without MEP subsidies. By all data series the down regulation value of wind power is higher in a system without MEP subsidy. However, these values express a potential value of wind power. Based on the current market regulation, the down regulation value of wind power is not accessible for a wind farm owner. The fourth row shows the short-term imbalance costs and the last row of the table expresses the imbalance costs as a percentage of total revenues on the APX. The imbalance costs of offshore wind power are significantly lower than the imbalance cost of onshore wind power. This is largely due to the ratio of imbalance volume over production volume. In the case of offshore wind power, there are more nominal production hours, which limit the forecast error and increases production volume. The figures in table 4.4 must be seen as preliminary results. Although we have used three data series of wind production, we only analyzed one year of data. To generate more solid results we need to analyze different years of wind power and market data. At this point we are not yet able to calculate the value of wind power. Only a few components have been quantified, based on a short-term perspective. Although valuing the price components from on an APX basis seems the most appropriate approach, it does not account for costs and values which are expressed in the long-term markets. Even when we succeed in quantifying all price components of wind power, they can not simply be added to- or subtracted from the wholesale electricity price to find the value of wind power, because there is a risk on double counting. The difficulty in isolating each individual price component still needs to be solved.
-57-
5
50680533-Consulting 07-0048
METHODOLOGY OF A MODEL
The aim of phase two of this study is to develop a model that gives the possibility to quantify the wind values and cost components for a given market configuration27 based on simulations and their resulting wholesale market price and based on balancing market behaviour (volatility and spread of the imbalance price). This chapter describes the methodology of that model and the perspectives that will be the starting point of the modelling activities. Furthermore, a schematic presentation of the model is given. The modelling will be done by chronologically simulating, on an hourly base, a perfect functioning electricity wholesale market assuming that each participant makes bids for prices equal to its power plant variable costs. The simulation can be leveraged, i.e. simulating a not perfect functioning wholesale market where participants exercise strategic behaviour in order to increase their profit (optimal equilibrium point is reached by maximum profit of each participant; based on supply function equilibrium theory of Klemperer-Meyer’s, see Appendix VII). In order to make the model suitable for different scopes, as for example: PPA28 (contract) negotiations, studies to request changes in regulation and market mechanisms, feasibility studies for investors, etc., we will first set up a methodology and define the steps to be considered each time a client29 wants to use this model for its defined scope.
5.1
Perspective of the PRP
As the client may be a wind farm owner or a utility, the first step30 is to define the perspective of the client, i.e. will the client take his own responsibility for the wind production imbalance by for example operating a ‘stand-alone wind farm’ or will this wind farm be part of a portfolio with more wind farms and more regulation capacity. Another perspective we propose to consider is the existence of one independent party in The Netherlands in charge of the Programme Responsibility services of all wind farms. In the following table these perspectives are summarized:
27
fuel prices, installed capacity, reserve margin, wind penetration, demand, etc. Power Purchase Agreement 29 refers in this paragraph to beneficiary parties of KEMA model (to be develop in phase two) 30 may be skiped if the scope relate to the perfect market 28
-58-
50680533-Consulting 07-0048
Table 5.1: Perspectives and strategies to commercialise wind power Perspectives
Portfolio
Strategy
Utility
Large (wind farms, flexible
Maximise profits
production plants, supplier) Wind farm owner operating its
1. One wind farm
1. Reducing risks
own portfolio
2. One wind farm plus one
2. Reducing risks
flexible power plant (gasfired)
3. Maximise profits
3. Several wind farms plus gasfired plant(s) National independent wind farm
All wind farms in The
Reducing risks and maximise
operator
Netherlands
profits
Each perspective in itself provides the information necessary to define the market structure (market participants and their portfolio’s) to be used as input for the simulation of strategic market behaviour. To this end we will use the mathematical model Symbad, which incorporate mathematic and economic theory, gaming theory and, last but not least, the learning process during years of simulation and experience of the electricity sector.
5.2
Scope of the simulation
The second step is to define the results what the client expects from the model, the period31 for which the model should apply and the assumptions and scenarios to be considered. Depending on the scope and the expected results (output data of the model), wind power’s value and cost components32 can be calculated based on simulation results such as: value of delivered wind power production33 during a considered period (e.g. year / month / hour), the wholesale market price on an hourly or even on a 15 minutes base, the capacity credits of a wind farm of a wind farm portfolio, the ‘Loss of Load Probability’, etc. Some other components can be calculated based on other methods, as the case may be for the wind pattern value, fuel diversification, geographical spread and administrative costs. The following table presents an overview of both methods to be used to calculate the value and cost components of wind power.
31
years in the future, present or past see previous chapter 33 hourly (15 minutes) wind production volume multiply with the hourly (15 minutes) wholesale price 32
-59-
50680533-Consulting 07-0048
Table 5.2: Methods to quantify value and costs components of wind power Value and cost components
ProSym simulations
Other methods (see former chapter)
Market value
X
Subsidy (MEP)
(X) X
Value components Image PRP (more customers) Wind pattern value
X X
Geographical spread
X X
Fuel diversification
X
Plant Margin Effect (capacity value)
X
Down regulating power
X
Cost components Short-term imbalance cost
X
Long term risks
X
Administrative costs
X
Based on insights of former market analyses we will define variables and scenarios to be used in ProSym simulation as part of the second step. To start with, the main variables should be: wind capacity penetration, cross-border capacity (Germany, Belgium, UK, Norwegian market), regulation (capacity payments, etc.), power plants (dismantlement, commission new plants), power plants flexibility, fuel prices, demand profile and volume, etc. As the next phase of this project aims to quantify wind value and cost components, we propose to use the model for the years 2005, 2010 and 2020. For the years 2010 and 2020 we will consider the scenarios presented in the next chapter based on several recent studies of KEMA, CBS, TenneT and UCTE. For the above mentioned years we intent to simulate all three perspectives, previous mentioned in step one. Choices should be made regarding the (number of) variables and the number of scenarios of each variable, depending on the considered scope(s).
-60-
5.3
50680533-Consulting 07-0048
ProSym and Symbad
The third step, which is the main part of the model, is formed by the two mathematical models mentioned in the previous section, i.e. ProSym and Symbad, which incorporate economic theory, simulations and experience of the electricity sector. In Appendix VI and Appendix VII descriptions of ProSym and Symbad are given. Based on the production simulation model ProSym and the strategic bidding model Symbad, the effect of offshore wind power penetration34 on the wholesale market can be investigated from different perspectives and for several scenarios. While the ProSym simulations are based on a perfect competitive market, Symbad simulates a more real market where market parties exercise strategic behaviour in order to maximise their profits. The main part of the simulation will take place in ProSym, while the different perspectives will be simulated by Symbad. The input variables are: - production plants and their technical data (see Appendix IV) - hourly import/export volumes and capacity - hourly demand - regulation changes, as: o capacity reserve contracts o introduction of capacity payments, etc. - for wind the following parameters will be considered: o hourly wind speed profile (based on historical hourly measured wind speed data during several years, in order to have consistent average and spread of the wind speed values, and the simultaneously effect with the hourly demand request and other hourly variables) o wind PV function, technical and non-technical availability o wind capacity penetration o flexibility and availability of the wind mills, and other technical data (as for example the technical improvements and development of the off-shore windmills technology) - granularity (hourly/15 minutes) - etc.
34
vary from 0 to 6,000 MW
-61-
50680533-Consulting 07-0048
Depending on the regulation scenario, Prosym can simulate the market not only for the SRMP (Short Run Marginal Price) but also for the LTMP (Long Term Marginal Price) or even for Forward (capacity) contracts. ProSym simulation delivers the hourly cost base price, the ‘Loss of Load probability’, the value of delivered wind power production and wind capacity credits as a function of wind power penetration. When also Symbad simulation is considered, this will adjust the ProSym cost base price with mark-ups determined from possible wholesale market prices based on optimum strategic behaviour of market participants. Some constrains will be presented in the form of available demand data and the simulation of embedded generation and CHP plants, for which assumption should be considered. The fourth and last step is the calculation of the value and costs components of wind power using the simulation results and the methods mentioned at step two (partly carried out in phase 1, see chapter 4).
-62-
50680533-Consulting 07-0048
Methodology
Knowledge
Input parameters
Matrix price components
Base case
Former We@Sea study
- LOLE - fuel prices (+CO2) - import/export - wind output - wind forecasterror Offshore wind - forecast-error - output profile - variability - availability
Inventory report Market - price volatility
Correlation analyses
- actual imbalance prices - contract prices on several markets (OTC, APX)
Scenarios
Market & Strategic Model
Production simulation Model
Stochastic model Arbitrage
Scenarios - development load profile - development interconnection capacity - development offshore windenergy capacity - development productiion mix - development fuel prices
Prosym Perfect market
Ouput - Price per MWh - capacity credit wind farm - regulating and reserve power
Risk bonus & value of wind Symbad Arbitrage Strategies (gaming theory) - imbalance minimisation - imbalance costs minimisation - risk minimisation - profit
-63-
6
50680533-Consulting 07-0048
SCENARIOS OF THE DEVELOPMENT OF THE ELECTRICITY MARKET
In this chapter we will present some figures of the base scenarios to be used for simulating the markets in 2005, 2010 and 2020. For the following variables the base scenario should be considered: - Technical data of production plants by type and unit - Installed wind power capacity (on-shore and offshore) - Cross-border capacities and volumes - Demand profile and volume. Besides the base scenario, scenarios will be considered for wind capacity penetration, wind speed profiles, wholesale prices in the connected markets (Germany, Belgium, UK, Nordpool market), regulation (capacity payments, etc.), fossil fuel prices (including CO2 emission credit prices), improvements of wind speed forecasts. Also variation of the base scenario could be considered. In the following tables the variables of the base scenario are shown. These figures have been derived from recent results of a We@Sea study, KEMA work-packages WP1 and WP2: “Balanshandhaving met 6 000 MW aan windenergie” (30450054-Consulting 2006-0123): Table 3
Energy Annual balance
2005 102 22 6 16 118
Base scenario (TWh) Electricity Production Imports Exports Trade Balance Demand (Including losses)
2010 117.9 23.5 7.5 16 133.9
Table 4 Demand and supply base scenario of 2020 Supply and demand by 2020 Base Yearly growth % 2% Demand volume
TWh
158
Production volume
TWh
137
Installed production capacity
GW
29,3
Growth production capacity reference year 2004)
GW
7,5
2020 144.8 27 9 18 162.8
-64-
Table 5
50680533-Consulting 07-0048
Capacity balance by 2020 (?)
Capacity balances Total Internal Net Generating Capacity Foreseeable not Available Capacity Connected Peak Demand Reserve Capacity
GW 24.13 1.21 23.6 4.72
Table 6 Electricity production by type (base scenario) Annual Electricity Production (TWh) 2005 2010 2020 Nuclear 3.7 3.7 0 Steam Thermal Units 38.3 45.8 44.7 Gas Turbine Units 0.8 0.7 0.7 Combined Cycle Units 50.8 57.8 87 Internal Combustion Units 1.6 1.6 1.8 Hydro 0.1 0.1 0.1 Non-fossil Renewables 2.9 3.8 4.9 Not Specified 3.8 4.3 5.6 Total 102 117.9 144.8 Generation growth scenarios For the growth of wind off-shore capacity we will refer to the steps as considered in the KEMA study Connect 6000 MW-II (40510025-TDC 05-48500) realized in 2005, i.e. 250 MW for each off-shore windmill park, up to a maximum capacity of 6 000 MW. The following scenario’s of on-shore and off-shore wind capacity to be considered.
Table 7
Installed wind power capacity
Installed capacity (MW) on-shore off-shore Total
2005 1200 0 1200
2010 1500 1180 2680
2020 1800 6000 7800
The modelling will include: Prosym: 1 perspectives * 1 strategies * 6 scenario’s * 2 years = 12 analyses Symbad: 4 perspectives * 1 strategies * 3 scenario’s * 1 year = 12 analyses
-65-
7
50680533-Consulting 07-0048
CONCLUSIONS
In this study the complex electricity market mechanisms, and especially the impact of (largescale offshore) wind power on these mechanisms, has been disentangled. The price of wind power has been decomposed in value and cost components and the influencing factors on these components have been determined. Some of the influences have been analyzed qualitatively others have been calculated quantatively. Especially the impact of wind power production on the APX price and on the imbalance volume have been studied in more detail. To this end we have applied statistical analyses and correlation techniques on one year of data (June 2004 – May 2005) consisting of wind speeds, APX prices and imbalance volumes and prices. We simulated the wind power production and earnings of both a 100 MW on-shore (on three different “locations”) and a 100 MW offshore wind farm. Only a few per cent of the variation in the APX price can be explained by variations in wind speed forecasts. During a few specific hours, mainly during the night in non-working days, occasionally 20-75% of the changes in the APX price is caused by wind power. This can be partly explained by the coexistence of relatively high wind speeds and low APX prices during nightly hours. Furthermore, during these hours the system switched over from “night load” to “day load” which means that the APX price was settled at the steep part of the market merit order. Thereby wind power production could lead to relatively large price reductions. During all other hours this effect is hardly recognizable. We found some similar results in the relationship between wind speed forecast error and system imbalance. The first conclusion is that a significant positive correlation is only shown occasionally between a short system balance – when electricity consumption is higher than the electricity consumption Programme Responsibility Parties have been forecasted – and a negative wind speed forecast error (more wind power has been produced than forecasted). The second conclusion is that significant positive correlations also is only shown occasionally between a long system balance – when electricity consumption is lower than the electricity consumption Programme Responsibility Parties been have forecasted – and the positive wind speed forecast error (more wind power was produced than forecasted). Most of these significant correlations have been found in sub series of nightly hours during non-working days, i.e. during hours with relatively low electricity demand. The overall conclusion is that, in general, the monthly and yearly correlations show that the wind power portfolio installed in 2004-2005 has almost no effect on the APX price and little effect on the TenneT imbalance market.
-66-
50680533-Consulting 07-0048
It is clear that wind power’s value is more than the average APX price only. Value components such as fuel diversification, wind pattern value and capacity value should be considered. Wind power also has the potential to provide down regulation power on the imbalance market. Optimally, a 100 MW wind farm owner could earn between 2.4 and 3.5 Euro/MWh for its balancing services when the MEP subsidies are included. Without the MEP there are no opportunity costs, which results in a down regulation value between 4.4 and 5.5 Euro/MWh. The short-term imbalance costs are the lowest for an offshore wind park with a weighted average of 2.2 Euro/MWh, while the short-term imbalance costs for an onshore wind park resulted in a weighted average price between 5.6 and 8.8 Euro/MWh. The lower imbalance costs of an offshore wind farm compared to an on-shore wind farms are mainly due to higher and less fluctuating wind speeds above sea. For all data series the onshore imbalance costs represented approximately 20% of total wind production income on the APX. For the offshore wind farm used in our case study the imbalance costs represents only 6% of the total income when the electricity produced would have been sold on the APX. The findings of this study contribute to the knowledge about wind power’s impact on the electricity market. Furthermore, the study is a first attempt to define and quantify the price drivers of wind power. But not all price components could be valued. Elaborated methods to quantify the price drivers of wind power require the development of a market model that calculates the wholesale market price for different scenarios. This model should also include strategic behaviour. Furthermore, the impact of wind power on the long-term electricity markets remains underexposed. A smart selection of financial derivatives could express the value of long-term risks. However, limited transparency of the long-term electricity markets is an important bottleneck. The aim of phase two of this study is to develop a model that gives the possibility to quantify the wind values and cost components for a given market configuration based on simulations and their resulting wholesale market price and based on balancing market behaviour (volatility and spread of the imbalance price). We already described the methodology of that model – based on ProSym and Symbad – and the perspectives that will be the starting point of the modelling activities. The modelling will be done by chronologically simulating, on an hourly base, a perfect functioning electricity wholesale market assuming that each participant makes bids for prices equal to its power plant variable costs. The simulation can be leveraged, i.e. simulating a not perfect functioning wholesale market where participants exercise strategic behaviour in order to increase their profit. We already drafted some scenarios for simulating the markets in 2005, 2010 and 2020.
-67-
50680533-Consulting 07-0048
REFERENCES Awerbuch S., Berger M. (2003) Applying portfolio theory to EU electricity planning and policy making. IEA/EET Working Paper. Report Number EET/2003/03. Paris. Bathurst, G.N., Grame et al. (2002) “Trading Wind Generation in Short Term Energy Markets,” IEEE Transactions on Power Systems, ieeexplore.ieee.org. Borchers H. (2005) “Plädoyer für marktwirtschaftliche Prinzipien bei der Abrechnung von Ausgleichsenergie” bne-kompass 02/05, pg. 4-10, Berlin. Cabral L.M.B (2000) Introduction to Industrial Organization, Massachusetts Institute of Technology. CEC, 2005 (Nakafuji, D-Y) Strategic value analysis – economics of wind energy in California. California Energy Commission, California Christensen H. C. (2003) General introduction to wind power in the Eltra area. Wind Conference, Billund, 20-22 October. DTe (2004) Onderzoek Ontwikkeling Liquiditeit Constateringen en aanbevelingen, Den Haag.
Elektriciteitsmarkt
2003
–
2004
DTe (2005a) Monitor methode groothandelsmarkt elektriciteit. Referentiedocument. Den Haag. DTe (2005b) Marktmonitor, ontwikkeling van de groothandelsmarkt voor elektriciteit 2004 – 2005 Resultaten en aanbevelingen Den Haag. ECN (2001) Energy market trends in the Netherlands, ECN policy studies. ECN, 2002 (Saint-Drenan, Y.M.) Wind power predictions analysis – Part.1, TenneT imbalance price system, development of a model for TenneT imbalance price. http://www.ecn.nl/library/reports/2002/i02010.html ECORYS, 2003. Presentation Liquidity Dutch Electricity Market. Electricity Supply Board, Ireland, 1990. CEC DG XII Wind Energy Penetration Study. Ensoc weekly (2006) Ensoc Energy Society. Nr.25 Juni 2006, Hilversum
-68-
50680533-Consulting 07-0048
ESB National Grid (2004). Impact of wind power generation in Ireland on the operation of conventional plant and the economic implications. ETSO, 2003. Current State of Balance Management in Europe. Balance Management Task Force EWI, 2003 (Lindenberger D., Schultz W.) Entwicklung der Kosten des ErneuerbareEnergien-Gesetzes; Kurzgutachten im Auftrag der Hydro Aluminium GmBH; Energiewirtschaftliches Institut an der Universität zu Köln; Köln, 10.01.2003. EZ, 2003 (Hoop Scheffer, J.N.G.). Steunmaatregel der Staten N 707/2002 – Nederland. Europese Commissie. Gjengedal, J (2003) “System control of large scale wind power by use of automatic frequency generation control (agc),”pp.15-21, 8-10 October. Giebel, G (2000) ‘A Variance Analysis of the Capacity Displaced by Wind Energy in Europe,’ Wind Power for the 21st Century, the Challenge of High Wind Power Penetration for the New Energy Markets, Kassel 25-27, pages 263-267. Hutting H.K., Cleijne J.W. (1999) The price of large scale offshore wind energy in a free electricity market. Proceedings of 1999 European Wind Energy Conference, Nice, pp. 399401 Holttinen H. (2003) Hourly wind power variations and their impact on the Nordic power system operation. PhD Thesis, Helsinki University of Technology. Holttinen H (2004) Impact of hourly wind power variations on the Nordic power system operation. VTT. Wind Energy. John Wiley & Sons, Ltd. ISET, 2000 (Ensslin, C., Hoppe-Kilpper M., Rohrig K.) WIND POWER INTEGRATION IN POWER PLANT SCHEDULING SCHEMES; EWEA 2000; http://www.iset.uni-kassel.de/abt/FB-I/publication/wind_power_plant.pdf ISET, 2001 (Kurt Rohrig, Dirk Christoffers) Prognoseverfahren zur optimalen Nutzung erneuerbarer Energien; FVS Themen 2001; http://www.fv-sonnenenergie.de/publikationen/th01/th2001_08rohrig.pdf
-69-
50680533-Consulting 07-0048
ISET, 2002 (Rohrig, K) Entwicklung eines Rechenmodells zur Windleistungsprognose für das Gebiet des Deutschen Verbundnetzes; ISF 2002, Berlin; http://www.iset.uni-kassel.de/abt/FB-I/publication/isf_02_ro.pdf ISET, 2003 (Ensslin C., Ernst B., Rohrig K., Schlögel F.) Online- Monitoring and Prediction of wind Power in German Transmission System Operation Centres; Hrsg.: EWEA, Madrid http://www.iset.uni-kassel.de/abt/FB-I/publication/EWEC_03_Be_En_HK_Ro.pdf KEMA, 2005 (Boer W.W.) Balanshandhaving Productnummer: 5033100 CL4. 05P4.5.1wwb
met
6000
MW
aan
windenergie.
Kollock P., Jaycobs R (2001) Liquidity Myths. Market Magazine Jan/ Feb Leonhard W., Müller K. (2002) Ausgleich von Windenergieschwankungen mit fossil befeuerten Kraftwerken – wo sind die Grenzen?; ew Jg. (2002) 101, Heft 21-22. Mothorst P.E. (2003) Wind Power and the Conditions at a Liberalized Power Market Wind Energy, Volume 6, Issue 3, pg. 297-308. Newbery, D., Fehr, von der N.H., Damme, van E (2003) Liquidity in the Dutch wholesale electricity market. NREL, 2004. (Parsons B., Milligan M., DeMeo E.A., Smith J.C.) Conference Paper. Wind Power Impacts on Electric Power System Operating Costs: Summary and Perspective on Work to Date. OSCOGEN, 2002 (Madlener, R., Kaufmann, M) Power exchange spot market trading in Europe: theoretical considerations and empirical evidence. European Union. Parsons B., Wan Y., Kirby B. (2001) Wind farm power fluctuations, ancillary services and system operating impact analysis activities in the United States. Proceedings of EWEC’01, July 2-6, 2001, Copenhagen; NREL report no. CP-500-30547 available at http://www.nrel.gov/publications. Parsons B., Milligan M., Zavadil R., Brooks D., Kirby B., Dragoon K., Caldwell J. (2003) Grid Impacts of Wind Power: A Summary of Recent Studies in the United States. European Wind Energy Conference, Madrid, Spain; June 2003.
-70-
50680533-Consulting 07-0048
Pinson, P et al.(2004) Optimizing the benefits from wind power participation in electricity markets using advanced tools for wind power forecasting and uncertainty assessment. Proceedings of the 2004 EWEC Conference, London, Nov. PJM MMU, 1999. Pennsylvania, New Jersey, Interconnection LLC. State of the Market Report.
Maryland
Market Monitoring
Unit
REVALUE 2, 2000 (Lorenzoni A., Fumagalli E., Morthorst P.E., Mitchell C.) The value of renewable electricity. Project Number: JOR3-CT98-0210 Sacharowitz S. Challenges and costs of integrating growing amounts of wind power capacity into the grid – some experience dealing with 12000 MW in Germany. Energy systems research group, Technical University Berlin. Berlin, Germany. SEI, 2004 (Milborrow, D. J.) Perspectives from Abroad. Assimilation of wind energy into the Irish electricity network. Soens, J (2005) Impact of wind energy in a future power grid. Katholieke Universiteit Leuven. Belgium Tauber C. (2002) Energie- und Volkswirtschaftliche Aspekte der Windenergienutzung in Deutschland, Sichtweise von E.ON Kraftwerke; Energiewirtschaftliche Tagesfragen 52. Jg. (2002) Heft 12. UCTE (2004) “ Integrating wind power in the European power system – prerequisites for successful and organic growth,” Tech.Rep. Ummels, B.C., Gibescu, M., Kling, W.L., Paap, G.C. (2004) Power system balancing with offshore wind power in the Netherlands, unpublished draft version. UWIG, 2003 (Smith J. C., DeMeo E. A., Parsons B., Milligan M.). Wind power impacts on electrical-power-system operating costs. Summary and perspective on work done to date, Presented at the American Wind Energy Association Global WindPower Conference, Illinois. Warren J., Hannah P., Hoskin R., Lindley D., and Musgrove P. (1995) Performance of wind farms in complex terrain. Proceedings of the 1995 British Wind Energy Association Conference, Warwick. Mechanical Engineering Publications Ltd, London
-71-
APPENDIX I
50680533-Consulting 07-0048
DEFINITIONS OF KEY TERMS
Reserve and regulation power Reserve and regulation power is offered on the imbalance market and traded in a mandatory system, where TenneT decides which bids and offers should be accepted to maintain system security. In general, fast starting production units, such as gas turbines, generate reserve and regulation power. In the general, the marginal production costs of reserve and regulation power are higher because the production units need to go through additional start and stop cycles, have little annual operational hours and use expensive fuels. Part-load PRP’s can decide to operate their power plants at partial load or "hot standby", which increases the load following capabilities, because the power plants can ramp-up their production level when needed. A limiting factor in operating power plants at partial load is the reduction in efficiency, which is undesirable from an economic and environmental perspective. Wind penetration level Unless otherwise stated, the wind penetration level refers to the (percentage) of wind energy delivered in a year, as a fraction of the total electricity consumption of the network concerned. As it is not always possible to derive this value, the parameter (wind capacity/peak demand) is used in place. Opportunity costs Opportunity costs are the costs associated with choices that lead to suboptimal profits. A wind park owner can miss revenues, when a positive forecast error occurs. In that case the wind power sold on the APX is less than the actual production. Moreover, the energy surplus needs to be balanced by TenneT and the price for spill energy can even become negative. This would imply that the wind park owner would have to pay for its wind production instead of making revenues. In the Dutch regulatory system a wind park owner can sell its electricity production against the market price plus the MEP subsidy. As the MEP subsidy for offshore wind power is currently three times the market price, it becomes clear that the opportunity costs of wind production are very high. Therefore, a wind park owner who optimizes its revenues will, in general, rather accept the risk on a negative forecast error than accepting the risk of a positive forecast error. This strategy could lead to a higher need for ramp-up power than ramp-down power.
-72-
50680533-Consulting 07-0048
In a formula, the opportunity costs can be expressed as: Opportunity costs = (Papx + MEP – Pspill) * (Eprogramme – Eactual) Imbalance ‘market’/system The imbalance ‘market’ is not a real market, because energy is traded on a mandatory basis, where TenneT is the only counterparty. However, the bids and offers for reserve and regulation power are accepted on basis of a market bid ladder. The imbalance ‘market’ clears every 15 minutes and is created to compensate all deviations between scheduled system balance (E-program) and actual real-time system balance. Imbalance volume Imbalance volume refers to the deviation between scheduled electricity production and actual energy production and is expressed in MWh. The imbalance volume can be determined on system level or for a specific party, such as a PRP or a wind park owner. Imbalance cost The imbalance costs are the costs needed to hedge or compensate the volume and price risk of wind power. Important to note is that imbalance costs of wind power are not only the costs TenneT charges to balance a deviation between scheduled and actual energy production, it also includes the impact on energy trading risks (market risk). The volume and price risks of wind power are further explained in section 3.2.3. Larger players will have relatively lower imbalance costs, because they can make use of a large generation portfolio to compensate the imbalance created by a specific production unit. Reserve and Regulation Supplier (RRS) A reserve and regulation supplier is a party which bids or offers dedicated regulation power on the supplemental energy market. To participate on this market a RRS needs to submit its bid or offer to ramp up production (reduce consumption) or to ramp down production (increase consumption) 1 hour before the period it is dedicated for. During each PTU, TenneT decides which bids or offers need to be accepted in order to maintain system balance. Active and passive players A distinction should be made between active and passive players on the imbalance market. Active players are reserve and regulation suppliers, which offer reserve power dedicated to keep the real-time balance, while passive players are PRP’s that can offer reserve and regulation power because their own balance position is negatively correlated (opposite) with
-73-
50680533-Consulting 07-0048
the balance position of the system. The reserve and regulation power of active players is thus guaranteed and submitted before the beginning of a PTU, while the reserve and regulation power of passive players depends on the balance position of PRPs during a PTU. The distinction between active and passive players can also be found back in the price difference between reserve power supplied by RRSs or PRPs Imbalance price model The price model determines the rules on which the imbalance prices are settled. In most countries the national TSOs use a one or a two price model to settle imbalance prices. One price model In a one price model one price is settled per PTU, which is indifferent of imbalance direction relative to the system imbalance. This means that the same price is used for ramp-up as well as ramp-down power. As a result, active regulating power does not need to be separated from passive balance, which gives a simpler settlement. This makes the pricing predictable and transparent and therefore there are less financial risks compared to a two price model. Two price model In a two price model different prices are charged for parties with imbalance in the same direction of the system balance and parties with an opposite direction of the system. Opposed to the one price model, a two price mechanism does make a distinction between passive and active contribution to balancing the system. The Dutch system is in principal a one-price model, because passive and active players are charged/paid the same price for reserve and regulation power. However, in 35% of the time TenneT charge two prices, when ramp-up and ramp-down power is needed at the same PTU (Saint-Drenan, 2002). Summarizing, the Dutch system applies a special kind of one price model, which has all the characteristics of a one price model combined with the ability to settle two prices per PTU. Capacity factor The capacity factor of a wind portfolio is the actual annual energy produced divided by the annual rated (nominal) output. The capacity factor of most wind parks varies between the 20-30% for onshore wind power and between 30-40% for offshore wind power.
-74-
50680533-Consulting 07-0048
Capacity credit The capacity credit is the percentage of the total installed wind capacity, which can be used to replace conventional capacity. The capacity credit of a wind power plant is based on its ability to deliver energy with the same loss of load as a conventional power plant. In general, the capacity credit lies between 20-30% of the total installed capacity. The capacity credit will increase with increasing wind power capacity, but not proportionally. The additional capacity credit will become smaller with every additional installed MW of wind power. This can be explained by the fact that at first the value of wind energy capacity will meet peak demand, however, when more wind capacity becomes available its relative contribution to meeting system load will become smaller. Seasonal effects influence the capacity credit Capacity value Apart from their value as energy producing entities, power plants have a capacity value. The capacity value of a power plant can be explained as the power plant’s contribution to system reliability. Plant margin The difference between the total installed reliable capacity and the maximal demand in a power system. Note that the plant margin appears to increase with wind on a network, due to the low load factor (usually termed "capacity factor") of the wind plant. E-program The E-program is a control mechanism of TenneT to balance the system. In this official document each PRP must state its physical position for t=0. The E-program needs to be submitted at t= -1,5, so that TenneT can check the E-program on internal and external consistency. PRP A program responsible party is an energy producer or an energy consumer which is obligated to submit their scheduled load or production in the E-program 1,5 hours before real-time. relative Production mix/ generation system The production mix is a term to describe all production units in the system. The Dutch production mix is mainly composed of gas-fired and coal power plants.
-75-
50680533-Consulting 07-0048
Ramp-up price The imbalance price paid to feed additional energy into the system or the price paid to reduce energy consumption Ramp-down price The imbalance price is paid to absorb additional energy from the system or to lower energy production Wind forecast A wind forecast is a prediction of wind power production levels in the future. With the current status of prediction tools the prediction window is not larger than 36 hours before t=0. Wind forecast error The difference between actual wind power production and forecasted wind power production is the wind forecast-error. With a positive wind forecast-error the PRP will be long and with a negative forecast-error the PRP will be short. A forecast-error will lead to imbalance costs, when TenneT needs to balance the deviation between scheduled and actual production. Wind power production Wind power production is defined as the amount of MWh fed into the grid by a wind power park for a certain time period. Being short or long A PRP will be short when it produces less energy than scheduled in its E-program. Equally, a PRP will be long when it produces more energy than scheduled in its E-program. Variability In this study the term variability is associated with the power output of a production unit and more specifically a wind power plant. As wind speeds are very dynamic, the wind power production level will change from minute-to-minute and from hour-to-hour. This stands in sharp contrast with conventional power plants which can deliver a constant power output level for long time periods. However, wind speeds above the wind speed needed for nominal production (16 m/s) will result in a constant output level, because of the variable speed technology in most modern wind turbines. Controllability Closely related to variability in power output, is the controllability of a power plant’s output. A power plant that can control its production level is able to respond on imbalance situations
-76-
50680533-Consulting 07-0048
and is thereby contributing to system stability. The controllability differs among the various production types which are present in the Dutch generation system. An often used indicator for controllability is the rate of change or ramp speed expressed in MW/minute. Gas-fired power plants generally have a high rate of change and are therefore able to respond quickly on imbalance situations. Coal power plants and especially nuclear plants have a relatively low rate of change. Wind turbine has very limited controllability. Technically they are able to ramp-down their power production when needed, but they cannot ramp-up their power production. Availability The availability depends on the frequency of forced outages (plant tripping or a storm front) and the amount of scheduled and unscheduled maintenance. The availability of power plants determines the long term balancing capabilities of the system. Although the technical availability of onshore wind turbines is rather high (98%), the long-term availability of wind power plants is quite uncertain, as they are dependent on unpredictable weather conditions. Price volatility Volatility is a measure for the deviation of a specific value (e.g. spot market price, wind energy yield) with a specific time horizon. When the price of electricity shows many price spikes and or dump prices the price volatility will be high. Higher price volatility implies higher risks, as it is much harder to predict the future electricity prices. Market liquidity Liquidity is a measure for the ease to find a trading partner for a given order on the market. The volume traded seems to be a good measure for the liquidity. Volatility needs to be considered as well. Market depth Market depth is a measure that reflects the size of an order needed to significantly move the market (i.e. the price), measured as the .price sensitivity for individual transactions Market transparency The market transparency indicates the degree at which market information is accessible. For a transparent market all participating parties must have equal access to information about volumes, prices, physical conditions of the system and many other factors to realize fair competition
-77-
50680533-Consulting 07-0048
Load Load is defined as the total demand for electricity Base-load Base-load describes the part of demand, which needs to be delivered at all times and shows very little fluctuations. Therefore base-load is mostly provided by power plants which have little controllability such as coal fired plants or nuclear plants. Peak-load Peak-load describes the part of demand, which is very time specific. During the day and the week there are peak-demand periods. However, peak-loads show far more fluctuation than base-load and are less predictable. Therefore peak-load is mostly provided by power plants with high controllability such as gas-fired plants. Ramp speed/ rate of change Ramp-speed or the rate of change describes the maximum change in production level, which can be achieved by a certain production unit. Firm capacity Firm capacity comprises all production units with a high controllability and availability of production Non-firm capacity Non-firm capacity comprises all production units with poor controllability and uncertain longterm availability. Marginal cost merit order In the market merit order all production units compete on their marginal production costs. The production unit with the lowest marginal production costs will be the first in the market merit order and will therefore have the best chance to sell its energy. Peak-performance capacity Peak-performance capacity comprises power plants with high ramp speeds. Interconnection capacity Interconnection capacity comprises the energy which can be exchanged between two or more different countries. In the Netherlands the import and export capacity is auctioned on
-78-
50680533-Consulting 07-0048
the APX market. Besides being used to meet the load, interconnection capacity is also used to balance the system. MEP In 2002 the Dutch Ministry of Economic Affairs adopted new legislation called ‘the Environment quality of the electricity production’ or MEP (EZ, 2003). The MEP provides renewable energy producers with a subsidy tied to production. Spot market/ spot price The Dutch spot market is the APX, at which spot prices are cleared for every hour of the next day. Intraday market An intraday market enables traders to trade between clearing of the spot market and the beginning of the imbalance market. On an intraday market traders can fine-tune their energy scheduling according the latest information. Especially wind power producers will benefit from an intraday market, as their forecast-errors will reduce with shorter prediction windows. Marginal costs In financial terms, the marginal costs are the variable costs per additional unit of output. “An important feature of wind energy is that the marginal costs are zero. In bidding on the market, wind energy can undercut any technology. Wind energy can therefore always be sold.” (Huttinge & Clijne, 1999). This is a major advantage in comparison towards fossil power plants. Considering that wind power is first in the market merit order, every additional megawatt of installed wind power will push fossil production capacity out of the market, at least for the times of wind availability. Load-following Load-following represents the efforts of the secondary control reserves to balance the system on an hourly time-scale. Load-following power is mostly provided by operation capacity, which can ramp-up or ramp-down its production level. Regulation power Regulation represents the efforts of the primary control reserves to stabilize system frequency on a minute-to-minute time-scale. Only operation power plants are able to regulate their production in order to stabilise the frequency of the power grid. All power plants are obligated to reserve a certain percentage of their nominal power output for primary control.
-79-
50680533-Consulting 07-0048
Balancing power Balancing power comprises all regulation and reserve power, which is dedicated to balance the system. Unit-commitment time-frame The unit-commitment time-frame represents the time-scale at which tertiary control reserves come into action. In general tertiary control reserves are power plants which have to make a black start, i.e., start production while the generator is cold. On average it will take 12 hours for a coal plant to produce at nominal power.
-80-
APPENDIX II
50680533-Consulting 07-0048
QUANTITATIVE CORRELATION ANALYSES
As (subsidized) wind power influences the spot and the imbalance wholesale market prices, we have tried to quantify this influence by analysing the following two correlations: - between APX price35 and wind speed forecast - between system balance36 and the (wind speed) forecast error37. Both correlations have been analysed for a one year period, i.e. June 2004 – May 2005. For the wind speeds and the wind speed forecasts, the following hourly data series where available: - Series 1: offshore (“average” location in the North Sea) - Series 2: on-shore (average value in the Netherlands) - Series 3: on-shore at meteorological station De Kooy - Series 4: on-shore at meteorological station Leeuwarden. The first two series have been used in a We@Sea project in which KEMA is involved (System integration and balancing preservation of 6,000 MW offshore wind); the latter two are data series from KNMI. We have used the data of series no.1 only to calculate the average prices on the APX price and the wind power price (see Chapter 4) as well as to calculate the imbalance costs of an offshore wind farm. The correlation analyses have been performed on series no. 2, 3 and 4.
II.1
Correlation analyse between APX price and wind speed forecast
Approach The research consists in analyzing the (negative) correlations between APX prices and wind speed forecasts. A negative correlation means that increasing wind speed forecasts resulted in a lower APX price. Results For the wind speed time series no. 2, 3 and 4 (on-shore locations) no significant yearly nor monthly correlation has been found between APX price and wind speed forecast.
35
Source: www.apxgroup.com Sum of imbalance volumes of all PV parties. Source: www.tennet.nl export data 37 Source: (wind speed) forecast error is defined as actual wind speed minus forecasted wind speed 36
-81-
50680533-Consulting 07-0048
To get more detailed results we segmented the data in months and further subdivided the days in five hour-blocks for working and for non-working days (so we have five times two is ten sub series of data for each month). The five hour-blocks have been defined based on price variation analyses as considered in some ECN studies. These hour-blocks are: - block 1: hour 00:01 to 06:00 - block 2: hour 06:01 to 10:00 - block 3: hour 10:01 to 16:00 - block 4: hour 16:01 to 20:00 - block 5: hour 20:01 to 00:00. In our analyses we concentrated on correlations (R) resulting in values of R2 > 0.2 which means that 20% of the changes in the APX price can be explained by wind speed forecasts. These values have been marked as “significant”. For the on-shore wind speed in the Netherlands (Series 2) only one sub series of data show significant negative correlations (correlation of R = –0.48 which means R2 = 0.23), i.e. the data of May 2005 for all night hours between hour 0 and 6 (block 1) on non-working days. As for all other blocks small negative and mainly positive correlations (with correlations up to 0.68 R2 = 0.46) have been found we conclude that occasionally significant negative correlations occur for Series 2 (about 2% of the time for the period June 2004 – May 2005). The analyses of the wind speed forecasts made for the meteorological stations De Kooy (Series 3) and Leeuwarden (Series 4) resulted in some higher values of the correlation between wind speed forecast and APX price. Negative correlations occur accidentally and vary up to R = –0.86 (means R2 = 0.74) for the sub series of February 2005 during nightly hours (block 1) on non-working days. Also during the same hours and days in January 2005 and during the evening (block 5) on non-working days in June 2004 we found significant negative correlations on both locations. These values can be partly explained by the coexistence of relatively high wind speeds and low APX prices during nightly hours. Furthermore, during these hours the system switched over from “night load” to “day load” which means that the APX price was settled at the steep part of the market merit order. Thereby wind power production could lead to relatively large price reductions. The frequency of negative correlations is roughly 10 times higher (17% of the time) compared to wind onshore in the Netherlands (Series 2). But overall the correlation between wind speed forecast and APX price for a specific location is very low.
-82-
50680533-Consulting 07-0048
In conclusion we can say that there is hardly any correlation between wind speed forecast and APX price. Significant negative correlations sometimes occur, mainly on non-working days during the night. In order to come up with more conclusions data series covering a longer period (a couple of years) and/or more detailed analyses of the existing data using further segmentation, as for different wind speeds and different APX price levels, is required.
II.2
Correlation between system balance and wind speed forecast error
For the wind speed forecast error38, the same data series have been used with actual measured wind speeds (see first page of this Appendix). The system balance39 and the onshore wind speed in the Netherlands (series 2) have been considered with a 15 minutes granularity. Unfortunately for wind speed data at meteorological stations De Kooy (series 3) and Leeuwarden (series 4) only hourly measured data were made available to KEMA. Approach In order to analyse the correlation between system balance and wind speed forecast error we filtered two situations: A. correlation between negative forecast error (less wind energy delivered than forecasted) and the “short position” of the system balance40 (positive system balance means consumption is higher than supply) B. correlation between wind speed with positive forecast error (more wind energy delivered than forecasted) and the “long position” of the system balance41 (negative system balance means consumption is lower than supply). The research consists in analyzing the positive correlations between system balance (short respectively long) and wind speed forecast error (negative respectively positive). Results No significant yearly correlation has been found for any of the time series, neither for the short system balance (situation A) nor for the long system balance (situation B). The same holds for the monthly correlations as the highest correlation found was for Series 2 in situation A (short system balance) with an R2 equal to 0.16 (February 2005 on working days). 38
difference between actual (measured) wind speed and forecasted wind speed sum of imbalance volumes of all PV parties in the Netherlands (for each 15 minutes) 40 to be addressed as short system balance 39
-83-
50680533-Consulting 07-0048
The following steps we have undertaken was to divide the day in five hour-blocks and consider this for working and for non-working days. The five hour-blocks we have defined for this part of the analyses are somewhat different compared to the blocks in the first section of this Appendix as they are based on price values and their standard deviation: - block 1: hour 00:01 to 02:00 - block 2: hour 02:01 to 06:00 - block 3: hour 06:01 to 09:00 - block 4: hour 09:01 to 17:00 - block 5: hour 17:01 to 24:00. We have calculated 60 correlations in total for non-working days in the period June 2004 – May 2005. The same number of correlations has been calculated for the working days, but with a different number of populations as there are two non-working days and five working days a week. In our analyses we concentrated on correlations (R) resulting in values of R2 > 0.2 which means that 20% of the changes in the system imbalance can be explained by wind speed forecast errors. These values have been marked as “significant”. We have found a few sub series in both situation A and in situation B showing a positive correlation resulting in R2 > 0.2. In the tables below an overview of these series is given.
Situation A Correlation results for the block segmentation between on-shore wind with negative forecast error and short system balance. Series 2 Non-working days
Working days
41
R2
Sub series (month, year)
Sub series (hours, block)
0.52
April 2005
between 2 and 6 (block 2)
0.36
June 2004
between 6 and 9 (block 3)
0.28
December 2004
between 17 and 24 (block 5)
0.53
February 2005
between 2 and 6 (block 2)
0.36
August 2004
between 17 and 24 (block 5)
0.28
February 2005
between 0 and 2 (block 1)
0.28
August 2004
between 2 and 6 (block 2)
0.25
October 2004
between 2 and 6 (block 2)
to be addressed as long system balance
-84-
Series 3
R2
50680533-Consulting 07-0048
Sub series (month, year)
Sub series (hours, block)
Non-working days
0.20
April 2005
between 6 and 9 (block 3)
Working days
n.a.
-
-
Sub series (month, year)
Sub series (hours, block)
0.25
September 2004
between 6 and 9 (block 3)
0.23
September 2004
between 2 and 6 (block 2)
0.20
October 2004
between 6 and 9 (block 3)
n.a.
-
-
Series 4 Non-working days
Working days
R2
Situation B Correlation results for the block segmentation between on-shore wind with positive forecast error and long system balance. Series 2 Non-working days
Working days
Series 3 Non-working days Working days Series 4 Non-working days
Working days
R2
Sub series (month, year)
Sub series (hours, block)
0.52
March 2005
between 2 and 6 (block 2)
0.35
January 2005
between 17 and 24 (block 5)
0.30
March 2005
between 17 and 24 (block 5)
0.28
September 2004
between 2 and 6 (block 2)
0.26
November 2004
between 2 and 6 (block 2)
0.25
April 2005
between 0 and 2 (block 1)
0.59
January 2005
between 17 and 24 (block 5)
0.38
January 2005
between 0 and 2 (block 1)
Sub series (month, year)
Sub series (hours, block)
0.23
February 2005
between 2 and 6 (block 2)
0.20
September 2004
between 0 and 2 (block 1)
n.a.
-
-
Sub series (month, year)
Sub series (hours, block)
0.34
November 2004
between 0 and 2 (block 1)
0.25
June 2004
between 17 and 24 (block 5)
0.25
September 2004
between 6 and 9 (block 3)
0.23
September 2004
between 0 and 2 (block 1)
n.a.
-
-
R2
R2
-85-
50680533-Consulting 07-0048
The conclusion is that a significant positive is shown occasionally between short system balance – when electricity consumption is higher than the electricity consumption Programme Responsibility Parties have been forecasted – and a negative wind speed forecast error (more wind power has been produced than forecasted). The frequency of occurrence of values R2 > 0.2 is about 3.3% of the time. Significant positive correlations also is shown occasionally between long system balance – when electricity consumption is lower than the electricity consumption Programme Responsibility Parties been have forecasted – and the positive wind speed forecast error (more wind power was produced than forecasted). For these situations the frequency of occurrence of values R2 > 0.2 is about 3.8% of the time. Most of the abovementioned correlations have been found in sub series of nightly hours during non-working days, i.e. during hours with relatively low electricity demand.
-86-
50680533-Consulting 07-0048
APPENDIX III CALCULATION METHODOLOGY CAPACITY CREDIT
Effective Load Carrying Capability: The best method for determining capacity value of intermittent generators is to calculate their effective load carrying capability (ELCC). This requires a reliability model that can calculate loss of load probability (LOLP), by using the loss of load expectation (LOLE), or expected unserved energy (EUE). ELCC is a way to measure a power plant’s capacity contributions based on its influence on overall system reliability LOLP: taking all such probabilities from each generator allows us to calculate the probability that enough generator units are on forced outage that the utility will be unable to meet its load. The primary advantage of a reliability-based assessment of capacity value is that it quantifies the risk of not supplying enough generation to meet loads. Intermittent renewable generators typically have low mechanical failure rates, but are not able to generate power when the resource is not available. Therefore, wind power generators have a higher LOLP. LOLE: The usual formulation is based on the hourly estimates of LOLP, and the LOLE is the sum of these probabilities, converted to the appropriate time scale. LOLE =
N i=1
P(Ci < Li )
where P() denotes the probability function, N is the number of hours in the year, Ci represents the available capacity in hour i, and Li is the hourly utility load. To calculate the additional reliability that results from adding intermittent generators, we can write LOLE' for the LOLE after renewable capacity is added to the system as: LOLE’ =
N i=1
P((Ci + gi) < Li )
where gi is the power output from the generator of interest during hour i. The ELCC of the generator is the additional system load that can be supplied at a specified level of risk (loss of load probability or loss of load expectation). ELCC wind integrated:
N i=1
P(Ci < Li ) =
N i=1
P((Ci + gi) < Li + Ci )
Calculating the ELCC of the renewable plant amounts to finding the values Ci that satisfies this equation. This equation says that the increase in capacity that results from adding a new generator can support Ci more MW of load at the same reliability level as the original load could be supplied (with Ci MW of capacity).
-87-
50680533-Consulting 07-0048
To determine the annual ELCC, we simply find the value Cp, where p is the hour of the year in which the system peak occurs after obtaining the values for Ci that satisfy the equation. However, intermittent plants also contribute to overall system reliability during nonpeak hours. Because LOLE is an increasing function of load, given a constant capacity, we can see from the equation that increasing values of Ci are associated with declining values of LOLE. Unfortunately, it is not possible to analytically solve the equation for Cp. The solution for Cp involves running the model for various test values of Cp until the equality in the equation is achieved to the desired accuracy. 1. Use load time series of base case system (conventional and other renewable energy sources) and calculate LOLE 2. Add power output serie of additional renewable generator (bv. 200MW) 3. Calculate LOLE’ and increase load until benchmark LOLE is achieved 4. Remove additional generator and calculate LOLE’, which will become higher than the benchmark LOLE 5. Add conventional power output serie until benchmark LOLE is reached (bv. 120 MW). In this step we can choose to use a gas or coal plant as equivalent for the capacity credit. 6. The amount of additional conventional power = ELCC 7. Capacity credit = ELCC/ installed capacity of renewable generator (120/200 * 100) This method comes down to depicting the LOLE curve of the gas equivalent and the LOLE curve of the renewable plant. Where both curves intersect will be the capacity that can be replaced by wind. The international benchmark figure for LOLE of conventional plants = 1 day/10 years, which corresponds to 2.4 hour a year. This reliability level is often used as a standard for utilities and provides a reasonable trade-off between cost and reliability. The capacity credit will increase with increasing wind energy capacity, closely related to the capacity factor, but not proportionally. The additional capacity credit will became smaller with every additional installed capacity of wind energy. At first the value of wind energy capacity will lie in meeting peak demand, however when more wind capacity becomes available it can only compensate for minor fluctuations of the system load and therefore it will have less value. If even more wind energy is added to the system it can not reduce the loss of load and can even increase the loss of load.
-88-
APPENDIX IV
50680533-Consulting 07-0048
TECHNICAL DATA OF A GENERATION UNIT - PROSYM INPUT DATA
Per generation unit the following input data is used: •
Station Name
•
Plant Name
•
Unit Classification (Must Run, Fixed, etc.)
•
Unit Efficiency Performance Factor
•
Start-up Model Parameters
•
Reserve Parameters
•
Minimum Up-time Constraints (Hours)
•
Minimum Down-time Constraints (Hours)
•
Start-up and Shutdown Ramping Constraints
•
Incremental Maintenance Cost
•
Unit Minimum Generating Level
•
Unit Maximum Generating Level (normal maximum)
•
Fuel Type
•
Fuel Constraints
•
Fuel Costs
•
Maintenance rates
•
Forced outage rates.
-89-
APPENDIX V
50680533-Consulting 07-0048
CALCULATING THE IMBALANCE COSTS RESULTING FROM A FORECAST ERROR
Table V.1 and V.2 illustrate how the imbalance costs of wind power have been calculated for a 100 MW offshore wind farm. Table V.1 shows how we used the measured wind speed data and the simulated forecast error of ECN, to calculate the forecasted wind speed per PTU. Hereby, we had to subtract the forecast error, corrected for nominal production wind speeds, by the actual wind speed. More information about the wind data can be found in the We@Sea (2006) study. In the next step we calculated the actual power production (MW) and forecasted power production by the following formula: P=
* Cp * A * v3
P=W = air density (kg/ m3) Cp = power coefficient A = rotor surface (m2) v = wind speed (m/s) The results of 4 PTU’s for series A are shown. In this project we calculated the values for all PTUs in the aforementioned period (June 2004 – May 2005) for all data series. The overall results are shown in chapter 4 of this report. Table V.1: Calculation of power production
PTU
Actual average wind speed (m/s)
Forecasted average wind speed (m/s)
Corrected forecast error
Actual power production (MW)
Forecasted power production (MW)
1
11,53
11,96
-0,43
75,0
83,9
2
11,59
11,67
-0,08
76,4
78,1
3
11,95
11,68
0,27
83,9
78,2
4
12,48
11,79
0,69
95,5
80,5
In table V.2 we calculated the wind production in MWh and subtracted actual production from forecasted production. The resulting forecast error in MWh represents the energy surplus or energy shortage of the wind farm. By multiplying the energy shortage of the wind farm with the imbalance price for ramp-up power, we got the imbalance costs in EUR/ MWh. In the four
-90-
50680533-Consulting 07-0048
PTUs we have used as example, there was only one imbalance price. There are also PTUs with two imbalance prices; one for ramp-up and one for ramp-down power. There are also PTUs where the imbalance prices become negative. A negative price for ramping-down means that a wind farm could actually be paid for its energy surplus.
Table V.2: Calculation of imbalance costs Productionactual (MWh)
Productionforecasted (MWh)
1
18,8
21,0
-2,2
30,0
2
19,1
19,5
-0,4
30,0
3
21,0
19,6
1,4
30,0
4
23,9
20,2
3,7
27,1
PTU
Forecast error (MWh)
Long price (EUR/ MWh)
Short price (EUR/ MWh)
Imbalance cost short (EUR)
Imbalance cost long (EUR)
30,0
-66,4
0
30,0
-13,1
0
30,0
0
42,3
27,1
0
101,3
Total
-79,5
143,6
Imbalance cost (EUR)
64,1
In this case the wind farm owner would have a net balance of 64,1 EUR after 4 PTUs. In the first and second PTU the wind farm owner would have to pay 79,5 EUR and in the third and fourth PTU, the wind farm owner would receive 143,6 EUR. However, the price for being long is generally lower than the spot price, which means the wind farm owner has to pay opportunity cost. In other words, although TenneT pays the wind farms owner for its energy production, he would probably have gotten more for his energy on the spot market. In this study we have not included opportunity costs in the imbalance costs.
-91-
APPENDIX VI
50680533-Consulting 07-0048
PRODUCTION SIMULATION TOOL - PROSYM
PROSYM is a chronological, multi-area electricity market simulation model that is often used to forecast electricity market prices, analyse market power, quantify production cost and fuel requirements, and estimate air emissions. It simulates system operation on an hourly basis by dispatching generating units each hour to meet load. The simulation is based on unitspecific information on the generating units in multiple interconnection areas (unit type and size, fuel type, heat rate curve, emission and outage rates, and operating limitations), and detailed data on power flows and transmission constraints within and between ISOs. Because the simulation is done in chronological order, actual constraints on system operation (such as unit ramp times, hourly spinning reserve constraints, and hourly transmission constraints and minimum up and down times) are taken into account. The resulting emission rates in one control region take into account emission changes in neighbouring regions. PROSYM has been used by many organizations, including the EPA and Department of Justice to pursue New Source Review violations, DOE, numerous utility companies, several ISOs and TSOs, Federal Energy Regulatory Commission (FERC), and Powering the South organization to simulate electric power system in the Southern U.S. The accuracy of the simulations depends on the composition of the system and the characteristics of the generating plants. For the Dutch system the inaccuracy of the average SMP calculation is estimated to be less than 2%. It is possible that the simulations don’t show the cheapest solution for every hour due to commitment logic. This means that the annual average market price as calculated with Prosym will most probably (>95%) lie between 102% and 100% of lowest possible price. This is of course based on the used assumptions. PROSYM simulation is used for daily operation, for short, medium and long term planning. Based on the existing infrastructure, the expected future power demand and the pre-selected future power units, power system configurations are composed to meet future demand. A power system configuration is a representation of a scheme of available power units and distribution capacity in each year of the evaluation period. Most of the characteristics may change every hour of the year. The model uses either marginal cost or bids to for its calculations. Power purchase contracts can be incorporated in the model as well as import and export of power. Transmission constraints and wheeling charges and are also evaluated. Furthermore PROSYM is capable of handling multi-area simulations, fuel contract variants, optimisation of multiple portfolio's and emission fees. The output of PROSYM can be in the form of a database, spreadsheet, print format or binary. Almost any variable can be represented on stations level. To give an example the illustration below gives the pool price together with the capacity output of the marginal unit on an hourly basis during a week (extracted from the bid based pool simulations for the validation of off shore wind energy).
-92-
50680533-Consulting 07-0048
Pool price and output marginal plant 120 100 80 60 40 20 0
200 150 100 50
pool price [$/MWh]
output marginal plant [MW]
week 27
0 1
25
49
73
97
121
145
Hour Marginal unit
Pool price
A representation of PROSYM and some of its modules and their relations is given in the illustration below. We however only used the PROSYM module in this study. Load data
LOADFARM
Load patterns
MULTISYM
Power system data
PROSYM
CHP
Run control
WATERWAY
TOPS
Output • in data base • in spreadsheet • formatted
-93-
APPENDIX VII
50680533-Consulting 07-0048
SYMBAD
Most electricity markets may be described as pools where output from different power plants is bid into a common power pool, i.e. the market. As the pool is bid-based, prices will not always reflect marginal cost since the presence of market power may induce strategic behaviour, or gaming. Therefore, it is widely accepted that game theory represents the most appropriate approach to describe the strategic behaviour of players in an oligopolistic market, such as the wholesale market for electricity. SYMBAD has been developed to be used in combination with the well known program PROSYM which is a tool to calculate a least-cost dispatch of a given range of power plants. Based on these results, SYMBAD simulates the bidding behaviour of specified generators, optimises the expected mark-ups of bid prices in the pool and predicts pool prices under different bidding strategies. To do so, the model relies on the supply function equilibrium which gives a set of strategies where no player has an incentive to deviate from his or her strategy. These mark-ups may then be used for a new simulation of all generating units participating in the electricity market. SYMBAD thus is a tool for estimating the mark-ups (both price and supply) due to strategic bidding of individual players in a bid-based power pool. These mark-ups reflect the extent to which the prices defined by the oligopolistic behaviour of competitors exceed perfectly competitive prices. The combination of PROSYM and SYMBAD offers a powerful new tool to assess the market, assist the development of an optimal bidding strategy or to make investment decisions into liberalised power markets.
Supply Function Equilibrium (SFE) Application in SYMBAD The SFE approach is applied in the gaming component of the SYMBAD algorithm and more precisely, in determining the bidding strategies of the players in the market. SYMBAD was built upon Klemperer-Meyer’s SFE theoretical framework that provides a necessary condition for reaching equilibrium. However, in order to define a unique SFE, we consider the linear version of the SFE that provides a unique solution. The marginal cost curves are approximated to linear or piecewise linear functions. The demand function is also linear, and incorporates the hourly peak load duration characteristic of the electricity markets, as well as a specific slope for each given demand segment.
-94-
50680533-Consulting 07-0048
Proofs and Validation Among the available techniques for estimating strategic behaviour of market competitors, KEMA has chosen the concept of SFE as the one that is considered to be the most applicable to wholesale electricity market conditions and specifics. This concept defines a set of equilibrium supply functions closed between Bertrand and Cournot extremes. These all are possible in reality, and are defined by the major factors specifying the particular market environment. These major factors are: market structure; degree to which market power may be exercised by some of the players; pool rules; degree of horizontal and vertical integration; degree of privatisation; etc. The various degrees and extent of influence of these factors on the market outcomes correspond to a particular equilibrium supply curve within the extremes. We need to assess the concrete equilibrium that occurs for a particular set of market conditions. In order to construct a working model that correctly reflects reality, we use the linear version of SFE. This requires a set of assumptions and limitations connected with this linear SFE to be defined. These are: affine or piecewise affine non-decreasing marginal cost curves; piecewise affine supply curves; and linear demand curves with constant slopes independent of time. Defining the curves in this manner enables calculation of unique SFE (unique mark-ups) while correctly and adequately approximating the real form of these functions. Exercises performed with the model showed that the results resemble the results of other models, and also showed that we are able to simulate actual occasions in a real market. The first example is a case where we modelled a European market in order to estimate future electricity prices. The figure below shows the results of the simulations in addition to the results of the client estimates.
-95-
50680533-Consulting 07-0048
80.00
SYMBAD Forecast
70.00 60.00
EURO / MWh
Client Forecast 50.00 40.00 30.00
SRMC
LRMC
20.00 10.00 -
2002
2003
2004
2005
2006
2007
2008
2009
2010
The results were comparable except for the years 2004, 2008 and 2009. After detailed examination, the client was convinced the Symbad results were more accurate. KEMA has completed projects on behalf of investment clients using Symbad to forecast revenues generated by power sector assets in Korea and Italy. Other projects with Symbad are the recently completed EU project “Analysis of the network capacities and possible congestion of the electricity transmission networks within the accession countries” and a study for market power assessment and mitigation in South Korea. We also mention the study performed for the Dutch regulator DTe: “Effects of regulatory measures on the Dutch wholesale electricity market” and the study performed for a Dutch electricity association: “Electricity price formation in relation to cross-border capacity”.
