System integration and balance control with large

System integration and balance control with large-scale offshore wind energy K. de Witte (Ecofys) S. van der Putten (Ecofys) H. Duivenvoorden (Ecofys) F. Nobel (Tennet) R. Beune (Tennet) M. Gibescu (TU Delft) B.C. Ummels (TU Delft) W.W. de Boer (KEMA) A.J. Brand (ECN) (We@Sea project 2004-010)

SYSTEM INTEGRATION AND BALANCE CONTROL WITH LARGE-SCALE OFFSHORE WIND ENERGY

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Ecofys Netherlands BV Kanaalweg 16-G P.O. Box 8408 NL- 3503 RK Utrecht The Netherlands T: +31 (0) 30 66 23 300 F: +31 (0) 30 66 23 301 E: [email protected] W: www.ecofys.com

SYSTEM INTEGRATION AND BALANCE CONTROL WITH LARGE-SCALE OFFSHORE WIND ENERGY

Ecofys: K. de Witte, S. van der Putten and H. Duivenvoorden TenneT: F. Nobel and R. Beune TU Delft: M. Gibescu and B.C. Ummels KEMA: W.W. de Boer ECN: A.J. Brand

Date: 31 March 2010

© Ecofys 2010 by order of: Stichting We@Sea

E COFYS N ETHERLANDS BV, A PRIVATE LIMITED LIABILITY COMPANY INCORPORATED UNDER THE LAWS OF T HE N ETHERLANDS HAVING ITS OFFICIAL SEAT AT U TRECHT AND REGISTERED WITH THE TRADE REGISTER OF THE C HAMBER OF C OMMERCE IN MIDDEN NEDERLAND UNDER FILE NUMBER 30161191

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Summary

Introduction This summary gives an overview of the results of the project: System Integration and Balance Control with Large-Scale Offshore Wind Energy (in Dutch: Systeemintegratie en balanshandhaving bij grootschalige windenergie op zee, “Si&Bh”). The project was conducted from 2004 to 2008 by TenneT, Ecofys, Delft University of Technology (TU Delft), ECN and KEMA under the We@Sea program. Tennet en Ecofys initiated the project in 2004. Objective In 2004 the Dutch government had set a target for the development of 6000MW of wind power in the Dutch part of the North Sea by 2020. The current system of balance control with its regulations (the so-called market design) was not designed specifically for the integration of a high quantity of wind energy. The market design may not be best suited for this integration. The objective of the project is to contribute to the design of an imbalance market in the Netherlands that can deal with large-scale wind energy. Results The project gives the following outcomes:  Insight in the flexibility and variability of electricity production on the basis of different scenarios for the development of (wind) power plants in the Netherlands until 2020 (workpackage 1 & 2);  Investigation of different energy storage technologies and fast start-up units an the possible synergies with wind power (workpackage 3). Work breakdown at the start of the project The work was broke down into 4 workpackages. In workpackage 1 (WP 1), various scenarios are set up for the development of wind power and conventional plants up to the year 2020. In WP 2, wind power variations on different time scales and their impact on balancing are investigated. WP 3 discusses the effectiveness of mitigating measures, such as fast start-up units, energy storage and short-term prediction of wind power. Also international experience in this area is summarized. Workpackage 4, we would put together the power scenarios, the potential impact of wind power on balancing and the mitigating measures, performing an integrated analysis of the imbalance that is to be expected and of the costs involved. Workpackage 4 wasn’t carry out.

i

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Outcome workpackage 1 & 2 In these workpackage firsts, we assessed the growth of conventional production and its flexibility, and define scenarios for the growth of wind energy, both up to the year 2020. Next, we explain how the wind speed data for simulating wind power production at system level are derived. On the basis of the growth scenarios, we then address the balancing challenges at system level due to wind power forecasting errors and also due to variations in the time scales within and larger than 15 minutes, up to 6 hours. By using a deterministic approach, typical situations and required control actions of Program Responsible Parties and the System Operator are identified. In addition, by using a stochastic approach overall statistics for the system balancing requirements are identified. Finally, we compare these requirements to the capabilities of the power system using the projected conventional generation park for 2020. In this summary we will give a short overview of the results of this workpackage. We will give insight in the flexibility and the variability of electricity production on the basis of different scenarios for the development of (wind) power plants in the Netherlands until the year 2020 and an assessment of the impact of large-scale wind power on the flexibility of conventional plants. For more detailed information of this worckpackage we refer to chapter 3 of this report. Flexibility of production is required to compensate unexpected wind power variations due to forecast errors. The flexibility of production, needed to follow the expected wind power variations, will most probably be sufficient. The 7.800 MW installed wind energy in the 2020 basic scenario will require an upward ramping ability from conventional generation of no more than 121 MW/min and a downward ramping ability of no more than 128 MW/min (see tbelow). Statistics of Wind Variability, Basic 2020 Scenario (7800 MW)

Time Range

Min (MW)

Max (MW)

99.7% C.I. (MW)

15 min.

-1810

1918

[-794.8 .. 852.8]

30 min.

-2398

2795

[-1200.0 .. 1221.8]

1 hour

-2683

3669

[-1807.8 .. 1744.2]

6 hours

-6742

6327

[-4678.0 .. 4619.6]

Given the expected maximal required ramp rate of 60 MW/min during the morning shoulder in 2020 the simultaneous required ramp rate could be (in a worst-case scenario) 190 MW/min. This is less than the 400 MW/min ramp rate estimated to be available from peaking gas units (maximum of 60 MW/min in 2020). The variations within 1 PTU due to turbulence do not result in significant additional control actions for the controller of TenneT (FVR), in case the wind energy produced during the PTU is adequately forecasted.

i

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

However, wind energy can result in scheduling problems in the case of high wind speeds at night. Base load units have to be shut down which is not economic. Solutions have to be found to handle this phenomenon. Due to the nature of wind, forecast is never perfect, which causes unexpected power imbalances if no counter-actions are taken after the scheduling based on the forecast 24 hours ahead. We have made a statistical analysis of these imbalances based on the entire one-year data of real wind speed measurements and forecast values. The imbalance power due to forecast errors 12-36 hours ahead varies between +3480 MW and -3900 MW for the 7.800 MW installed wind power in the 2020 Basic scenario, given a 99.7% confidence interval. This means that the extremes will not be exceeded for more than 1.1 days/year. This requires between 45% down-regulation and 50% up-regulation or reserve power (percentages with respect to the 7800 MW installed capacity). The up-regulation amount is about 5.5 times bigger than the current N-1 reserve requirement (~ 700 MW), set for the largest conventional unit outage in the Netherlands. It can be concluded that the impact of large amounts of wind energy needs additional reserve requirements to deal with additional variability at the system level. Outcome workpackage 3 Fast start-up units and Compressed Air Energy Storage The variability of large-scale wind power requires additional flexibility from the rest of the generation system. The combination of wind power generation with energy storage plants is frequently regarded as an optimal combination to deal with the variability of wind power. Also, fast start-up units may provide interesting opportunities. In this workpakage, different energy storage technologies and fast start-up units are investigated and conclusions are drawn regarding possible synergies with wind power. Dedicated capacity for balancing wind power First, the effectiveness of various storage technologies and fast-start up units are assessed on a back-to-back basis with wind power. This is done using the forecast power imbalances of wind power obtained in WP2. The investigated technologies are a 3000 MW, 10 GWh pumped storage system in combination with 852 MW of fast-start up units (open cycle gas turbine, OCGT). In the figure, the use of energy storage and fast start-up units decreases the overall power variations: in case wind power falls short of the prediction, generation of energy storage and OCGT can be used, in case wind power exceeds the predicted value, energy storage is used to take in the additional wind power. Clearly, the impacts of wind power forecast errors are reduced by this configuration.

ii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

When balancing wind power, it must be taken into account negative power imbalances (wind power falls short) has different impacts on the system than positive power imbalances. In case wind power falls short, additional generation is needed and units with fast ramp-up capabilities or short start-up times are required. In case wind power is long, generation levels of units in operation will need to be reduced, requiring flexibility with respect to ramp-down or short shut-down capabilities. In the report, negative power imbalances have been considered in more detail for a number of combinations of energy storage and fast start-up unit capacities.

Forecast Imbalance (MW)

The larger the reservoir of energy storage, the larger the time energy storage will be able to back-up negative wind power forecast deviations. It is found that in order to arrive at a system reliability comparable to present levels at least a installed capacity of 2.6 GW of fast start-up units is required to balancing wind power shortfalls, as well as a 10 GWh reservoir for energy storage. On average, these units would be operating only 23% of the time. It should be noted that these number partly result from the fact that wind power is balanced by dedicated capacity; would a more system-wide approach be applied (wind power imbalance as part of the total system imbalance), different results would be obtained.

4000 Original

2000

Reduced

0 -2000 -4000 1

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

1.5 4

10000

5000

0 Fast-Startup Power (MW)

Storage Energy (MWh)

x 10

1

1.05

1.1

1.15

1.2

1.25 1.3 PTU Number

1.35

1.25 1.3 PTU Number

1.35

1.4

1.45

1.5 4

x 10

800 600 400 200 0

1

1.05

1.1

1.15

1.2

1.4

1.45

1.5 4

x 10

Overall power deviations due to forecast errors, energy storage reservoir level and operating points of OCGT for a simulated time-series (reference Annex 7)

iii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Intelligent charging/discharging strategies Two installed pumped storage systems (2 GW and 3 GW installed capacity, reservoir size 30 GWh) are compared for an intelligent control strategy. This strategy always aims at returning the storage system to a half-charged state (15 GWh), thereby allowing maximum flexibility with respect to the wind power forecast (energy storage may be used both for generation and storage at all times). Furthermore, a systemwide approach mentioned above is applied: the pumped storage is operated on a market-basis, making profits from the differences in peak and off-peak prices. Obviously, the high level of wind power examined in this research can be expected to have an impact on power prices on the market: high winds leads to low prices and vice versa. Opportunities for Compressed Air Energy Storage (CAES) From the analyses performed for the opportunities for compressed Air Energy Storage (CAES), it can be concluded that the technology for diabatic CAES systems is available and has been applied successfully in a small number of cases. It is noted that many alternatives exist (conventional generation ramping, incidental wind power curtailment, extension of network capacity). Also, the profits of CAES will be difficult to assess, because these heavily depend on future power prices and the market design. It is concluded therefore that without dedicated support, CAES development in Europe will be restricted to a very limited number of initiatives (e.g. at already existing, unused salt domes). Whether such dedicated support may be justified by policy targets related to wind energy is still questionable. Day-ahead wind power forecasts Improved wind power forecasting reduces wind power forecast errors and its associated power imbalances, and thereby facilitates a more economical integration of wind power into power systems.. This paragraph provides insight into the accuracy of present forecasting methods and the opportunities of improvements. Also, the effect of minimizing wind power forecast errors by aggregation of geographically spread wind power is considered. Wind power variability and prediction errors The value of wind energy for the Dutch electricity supply has been assess using one year of 15-minute wind speed time series for 7.8 GW wind power (Basic Growth 2020 Senario). It was found that the capacity factor was around 48%, with monthly means varying between 35% (summer) and over 50% (winter). Wind power production variations were found to vary between 0 an 2,2 GW/15 min, with variations between 1.5 and 2.2 GW/15 min. occurring twenty times per year, mostly due to simultaneous cut-out or cut-in because of storms or calms. Clearly, the spatial smoothing effect limits the power variations of the total installed wind power capacity.

i

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

In order to investigate the impact of prediction age and imperfect predictions on wind power balance, the same method has been used to calculate the 15-minute predicted wind power for four different day-ahead forecasts (24h, 18h, 12h and 6h ahead). The accuracy of a prediction is evaluated by comparing the predicted values to the actual produced amounts. The forecast error of the power prediction is expressed by the Normalized Mean Absolute Error (NMAE). For the 24h. Ahead forecast, NMAE was found to have an average of 840 MW, while 70% of the forecast errors are within a 730 MW range, or 10% of installed capacity. Influence of Forecast Lag on the Required Regulating Energy By analysing the four different day-ahead forecasts, it can be shown that the impacts of bad predictions can be avoided by making use of forecast updates, reducing the average prediction error to 720 MW using the 6 hour forecast. This clearly shows the importance of continuous wind power forecast updates, which also allow a better allocation of the forecast errors within the operation of other generation units in the system. It is shown that the forecast error's normalized standard deviation drops to half from the 36 to the 2-3 hours ahead prediction. Interestingly, the analysis also shows that predicted wind power variations are smaller than the actually realised wind power variations, which is especially for large wind power variations. Obviously, the latter conclusion depends on many factors and may be different for other wind speed data years. Furthermore, it should be noted that NMAE of NWP-based forecasts does not reduce to zero when the forecast horizon is 6 hours due to the intrinsic uncertainty in NWP models. Such a reduction however can be achieved if online production data is included in the forecasts, as is done in Error! Reference source not found., showing an example of the normalized standard deviation of the prediction error for the 0-36 hours ahead wind power forecast for 8000 MW capacity.

Forecast error standard deviation for 8000 MW wind power and forecast horizon (reference Annex 8)

ii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Aggregation of forecast errors and power imbalances This section specifies the impact of the aggregation of large-scale wind power on the overall forecast error and power imbalance due to wind power forecast errors. Two aggregation levels are considered: at the systems level and at the programme responsible party (PRP) level, being 7 individual market parties each with some wind power as part of their portfolio. The hypothesis was that a central aggregation of power imbalances would allow some internal canceling out of wind power variations. It is found that this is indeed the case: aggregation at the system level requires less 5– 15% overall reserves for the balancing of wind power variations and the compensation of forecast errors (table below). Power variations and forecast power variations for individual PRPs and Systems

15-min variation

forecast error

Max. Min. [MW/PTU] [MW/PTU]

99.7% C.I. [MW/PTU]

Max. [MW]

Min. [MW/PTU]

99.7% C.I. [MW]

Sum PRPs

+1915

-2294

[-1030 ... 969]

+5257

-5450

[-3718 ... 3972]

System

+1810

-1980

[-853 ... 799]

5148

5326

[-3482 ... 3907]

Difference

105

314

-177 ... 170

109

124

[-236 ... 65]

Wind Power Control Strategies Offshore wind farm control strategies may reduce the power variations of wind power. For all control measures it applies, that the wind power output can never exceed the available wind power. Thus, wind power may be reduced at all times, but for upward control actions, an initial derease of wind farm output may be required. It can be noted that the design of the wind turbine, in particular the presence or absence of a cut-out wind speed, has an important impact on the opportunities for wind farm control. Concluding, it should be borne in mind that any wind farm control strategy brings about some opportunity loss of wind power production. The extent to which such strategies may be regarded as optimal therefore heaviliy depends on the market design for wind power, subsidy schemes and other, non-technical factors. Control Strategies for Dealing with Large Power Variations A number of technical possibilities exist for controling the wind park output, which are shown in Error! Reference source not found.. The possibilities include an absolute output level limitation, ramping for balance control, ramp rate limitation (available only up to a certain extent for downward regulation) and delta control (fixed operation percentage of maximum available).

i

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Wind farm control strategies applied at an existing offshore wind farm (reference Annex 4)

An estimate has been made of the benefits of control actions for the integration of large-scale wind power, using the wind power output time-series obtained in WP2. Furthermore, the predictability of large wind power variations is estimated with and without the application of these control strategies. The regulating strategies assessed here investigate the limiting of ramp rates and simultaneous cut-out in particular. The results show that ramp rate limitations indeed enable a reduction of the maximum wind power output variations. Also the possibility of ‘storm control’, in which a gradual shut down of the turbine is assumed as compared to a sudden turbine cut-out, allows a reduction of large wind power variations in number and in size. It is shown that the possibilities of both strategies depend on the maximum power variation level deemed acceptable. For limiting the down ramp, it may be necessary to reduce wind power output as a precaution as well: predicted variations may not happen, and variations may be different then predicted. Bidding and Down-Regulating Strategies This section combines the wind power forecast and wind power prediction data (WP2) with with price information from different markets. Price data have been obtained from the day-ahead spot market operated by the Amsterdam Power Exchange (APX) and from the hour-ahead single-buyer imbalance market operated by the Dutch TSO TenneT. Combining these data allows the computation of spot-market revenues and imbalance penalties or gains for large-scale wind power. Three different cases are evaluated:   

Perfect wind power forecast (zero prediction error); Median bid (realistic prediction error, forecast median bid into spot-market); Optimal quantile bid (realistic prediction error, forecast median bid into spotmarket).

The optimal quantile is determined from the probability distribution function of the random 15-minute average wind power production, and depends on price predictions for the APX day-ahead and TenneT imbalance prices. The wind producer may use two down-regulating strategies to avoid being charged for excess wind. The yearly revenues for various combinations of bidding and down-regulating strategies are then computed for an example wind producer. The results are shown in the table below.

ii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Revenues for different bidding strategies

From the results, it can be concluded that bidding the optimal quantile combined with down-regulating to zero output results in the largest yearly revenues. It is interesting to note that for this data set, the optimal quantile bidding combined with a strategy of down-regulating to contract level produces a comparable income with that of an equivalent producer with a perfect wind power forecast.

i

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Table of contents

1

Introduction ............................................................................................. 1

1.1

General introduction.............................................................................. 1

1.2

Objective of the project ......................................................................... 1

1.3

Relation with the We@Sea program......................................................... 1

1.4

Expected results of this project ............................................................... 2

1.5

Project strategy at the start of the project ................................................ 2

1.6

Overview of this report .......................................................................... 3

2

Work breakdown and partner tasks.......................................................... 4

2.1

Workpackage 1: Scenarios for the electricity sector ................................... 6

2.2

Workpackage 2: Variations as consequence of wind energy integration......... 6

2.3

Workpackage 3: Instruments for balance control ....................................... 7

2.4

Workpackage 4: Analysis and solutions .................................................... 8

3

Summary results workpackages 1 and 2 ‘Scenarios and analysis of current methods for balancing’................................................................. 9

3.1

Abstract .............................................................................................. 9

3.2

Introduction ....................................................................................... 10

3.2.1

Overview of this executive summary ..................................................... 10

3.3

Scenarios for the electricity sector......................................................... 12

3.3.1

Description and growth of other plant .................................................... 12

3.3.2

Flexibility of production........................................................................ 13

3.3.2.1 Growth of wind energy ........................................................................ 13 3.4

Time series of measured and forecasted wind speed ................................ 15

3.4.1

Methodology ...................................................................................... 15

3.4.2

Time series of measured wind speed ..................................................... 16

3.4.2.1 Transformation to turbine hub height..................................................... 16 3.4.2.2 Identification of daily patterns and interpolation to wind farm locations ...... 17 3.4.2.3 Identification of random variations and interpolation to wind farm locations 18 3.4.2.4 Transformation to 15 minutes averaging time ......................................... 19 3.4.3

Time series of forecasted wind speed..................................................... 20

ii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.5

Variations at different scales and control response - Deterministic approach 22

3.5.1

Deterministic approach ........................................................................ 22

3.5.2

Power variations within 1 PTU............................................................... 22

3.5.2.1 Methodology ...................................................................................... 22 3.5.2.2 Power variations due to normal wind speed variations.............................. 22 3.5.2.3 Power variations due to strongly correlated “gusts” ................................. 24 3.5.3

Power imbalance over more than 1 PTU ................................................. 25

3.6

Variability and predictability of wind power production - Stochastic approach27

3.6.1

Methodology ...................................................................................... 27

3.6.2

Statistics of Wind Variability at Different Time Scales ............................... 29

3.6.3

Statistics of Wind Predictability ............................................................. 30

3.7

Summary of required and available flexibility of production....................... 33

3.8

References WP 1&2............................................................................. 34

4

Summary of Workpackage 3 ‘Instruments for Balance Control’.............. 35

4.1

Introduction ....................................................................................... 35

4.2

Overview of this summary ................................................................... 35

4.3

Fast start-up units and Compressed Air Energy Storage (KEMA, TU Delft, Ecofys).............................................................................................. 36

4.3.1

Dedicated capacity for balancing wind power .......................................... 36

4.3.2

Intelligent charging/discharging strategies ............................................. 37

4.3.3

Opportunities for Compressed Air Energy Storage (CAES)......................... 38

4.4

Day-ahead wind power forecasts (Ecofys, ECN, TU Delft).......................... 38

4.4.1

Wind power variability and prediction errors ........................................... 38

4.4.2

Influence of Forecast Lag on the Required Regulating Energy .................... 39

4.4.3

Aggregation of forecast errors and power imbalances............................... 39

4.5

Wind Power Control Strategies (ECN, ECOFYS, KEMA, TU Delft)................. 40

4.5.1

Control Strategies for Dealing with Large Power Variations ....................... 40

4.5.2

Bidding and Down-Regulating Strategies ................................................ 41

4.6

International Experiences (Ecofys, KEMA)............................................... 42

4.6.1

Balancing markets .............................................................................. 42

4.6.2

Literature survey ................................................................................ 43

iii

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

5 5.1

Workpackage 4: Analysis and solutions .................................................. 45 International communication about the project results ............................. 45

iv

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

1

Introduction

1.1

General introduction

This document gives an overview of the results of the project: System Integration and Balance Control with Large-Scale Offshore Wind Energy (in Dutch: Systeemintegratie en balanshandhaving bij grootschalige windenergie op zee, “Si&Bh”). The project was conducted from 2004 to 2008 by TenneT, Ecofys, Delft University of Technology (TU Delft), ECN and KEMA under the We@Sea program. Tennet en Ecofys initiated the project in 2004.

1.2

Objective of the project

In 2004 the Dutch government had set a target for the development of 6000MW of wind power in the Dutch part of the North Sea by 2020. The current system of balance control with its regulations (the so-called market design) was not designed specifically for the integration of a high quantity of wind energy. The market design may not be best suited for this integration. The objective of the project is to contribute to the design of an imbalance market in the Netherlands that can deal with large-scale wind energy.

1.3

Relation with the We@Sea program

In order to meet the 6000MW target, knowledge and technical expertise are required to develop and operate wind farms in a sustainable way. The provision of a subsidy to gain expertise and knowledge for this goal was the driving force in the formation of the consortium We@Sea. The aim of We@Sea is to gather knowledge and reduce risks, in order to realize a sound implementation of wind energy in the North Sea. The first two Dutch offshore wind farms are in operation -- Noordzee Wind (2006) and Prinses Amalia Wind Park (2008) -- and new farms are being developed. The experiences of these first projects will be used in the development of new projects. The application of acquired knowledge and experience remains a continuous process. We@Sea plays an active role in that process. This project connects directly to the work defined in workpackage 3.4 of the We@Sea program: System analyses. The objective of the project is to determine the properties of a reliable electrical supply system with large scale offshore wind energy integration. The system must maintain the same level of reliability as in the current situation, taking into account the fluctuating nature of wind power generation and consumption, both at short time scales (seconds to minutes) and long time scales (minutes to days).

|1

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

1.4

Expected results of this project

The expected results of the project Si&Bh are: 

 

   

Insight in the flexibility and variability of electricity production on the basis of different scenarios for the development of (wind) power plants in the Netherlands until 2020 (the results can be find in the conclusions of Chapter 3); Assessment of the impact of large-scale wind power on the flexibility of other plants (the results can be find in the conclusions Chapter 3); Identification of mitigating measures, their potential to reach the required flexibility and the collateral effects (e.g. cost, environment) (the results can be find in paragraph 4.1 – 4.5); An overview of international experience with high penetration of wind energy (the results can be find in paragraph 4.6); Insight in the cost of imbalance due to wind power (wasn’t carry out); Methods for risk (e.g. cost) management in system balancing (wasn’t carry out; Recommendations and arguments to be used in a discussion on the design of an imbalance market in the Netherlands which deals with large-scale wind power (wasn’t carry out).

The project has been carried out into four workpackages. Chapter 2 gives an overview of the tasks of the different workpackages.

1.5

Project strategy at the start of the project

The integration of large scale wind power will possibly have far-reaching consequences for the development of (offshore) wind energy but also for other plants. Given the long term nature of these developments (and the related investment), it is crucial that the various players in the market bring in their interests at an early stage. By participating in the project, project partners actively contribute to the future market design. The project does not aim necessarily to reach consensus between stakeholders on the construction of a future market design. However, identification of discrepancies and possible solutions is a prerequisite for an effective political discussion that then must be run. The project will not only substantially contribute to survey and to remove the potential significant barriers to 6000 MW in 2020, but the project will also lead to a number of intermediate results that may be of interest to the program responsible parties.

|2

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

1.6

Overview of this report

This final report of the Si&Bh project summarizes the results of the different workpackages.    

Chapter 2 presents Chapter 3 presents Chapter 4 presents Chapter 5 presents of the international

the work break down and the partner tasks of this project; a summary of the results of workpackage 1&2; a summary of the results of workpackage 3; a summary of the results of workpackage 4 and an overview dissemination of the project results.

The results of the studies that are carried out in the context of this project are enclosed in the Annex Report of this project.

|3

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

2

Work breakdown and partner tasks

The work in the project is broken down into various workpackages, activities and studies and is distributed over Ecofys (applicant and project coordinator), TenneT (applicant), Delft University of Technology (TU Delft), ECN and KEMA; see Table 1. Work breakdown at the start of the project In workpackage 1 (WP 1), various scenarios are set up for the development of wind power and conventional plants up to the year 2020. In WP 2, wind power variations on different time scales and their impact on balancing are investigated. WP 3 discusses the effectiveness of mitigating measures, such as fast start-up units, energy storage, and short-term prediction of wind power and wind farm control strategies. Also international experience in this area is summarized. Finally, the power scenarios, the potential impact of wind power on balancing and the mitigating measures are put together in WP 4, performing an integrated analysis of the imbalance that is to be expected and of the costs involved. The outcome of this effort will indicate if and to what extent the current design of the imbalance market in the Netherlands has to be modified.

|4

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Table 1 Initial work break down WP 1: Scenarios electricity sector a.

Identify current generating

KEMA, TUDelft

b.

Develop scenarios regarding production until 2020 Develop scenarios regarding wind power development until 2020

KEMA, TUDelft

c.

TUDelft, KEMA

WP 2: Variations as consequence of wind energy integration a. Identify wind power variations within 1 ECN, TenneT PTU b. Identify wind power variations between ECN, TenneT PTU’s separated 15 minutes c. d. e.

Identify long-term wind power variations Impact on balancing (or maintaining the balance)

ECN, TenneT, Ecofys KEMA, TUDelft, TenneT

Research utility capabilities and energy predictor (optional)

ECN

WP 3: Resources for balance control a.

Fast start-up generation units;

KEMA, TUDelft, TenneT

b.

Windpower storage technologies;

KEMA, TUDelft

c.

Day-ahead wind power forecasts

ECN, TenneT

d.

Optimizing park layout

ECN, Ecofys

e.

Wind farm control strategies (including cost impacts);

ECN, Ecofys, KEMA, TUD, TenneT

f.

International experiences.

KEMA, Ecofys, TenneT

WP 4: Analysis and possible way forward a. Degree of matching

KEMA, ECN, TenneT KEMA, ECN, TenneT KEMA, ECN, TenneT KEMA, ECN, TenneT, TUD

b.

Costs of balance control

c.

Regulation modification?

d.

Possible way forward and solutions

e.

Recommendations

KEMA, ECN

Management and coordination

Ecofys

|5

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

2.1

Workpackage 1: Scenarios for the electricity sector

Based on information of the sector, a number of scenarios were prepared. These scenarios show the possible composition of the generation mix in 2020. The development of exchanges with foreign countries is included (autonomous development). Also taken into account is growth of wind power in neighboring countries and the implications this potential for balance sheet management in the Netherlands may have. For the growth of wind power in the Netherlands different scenarios are prepared based on consultations with the market (link to RL0 We@Sea). The following activities have been taken place under WP 1: a. identify current generating b. develop scenarios regarding production until 2020 c. develop scenarios regarding wind power development until 2020 The results of workpackage 1 are summarized in chapter 3 (paragraph 3.4). In Annex 1 and Annex 2 of the Annex Report the extensive research results of the sub-studies under this workpackage can be found. 2.2

Workpackage 2: Variations as consequence of wind energy integration

In WP 2, we investigate wind power variations on different time scales and their impact on balancing. Results of previous studies will be integrated. The following activities take place under WP 2: a. Identify wind power variations within 1 PTU b. Identify wind power variations between PTU’s separated 15 minutes c. Identify long-term wind power variations d. Impact on maintaining the balance In contrast to what the project breakdown (table 1) suggests, ECN's work in WP 2a, 2b and 2c did not consist of determining wind power variations but consisted of creating wind input for the simulations (performed by KEMA and Delft University of Technology) that gave the wind power variations. These wind data include measurements as well as forecasts. As a consequence data to be employed in WP 3c (scheduled for 2006) became available by December 2005. In this context it also must be mentioned that the work of Delft University of Technology included the creation of time series of measured and forecasted wind speed in the future wind farm locations out of such series in locations where wind was measured. With regard to the topics of WP 2 the initial division in 3 time spans (within 1 PTU, within 1 hour, and longer than 1 hour) was changed into 2 time spans (within 1 PTU, and between PTU's separated by 15 minutes, 30 minutes, 1 hour and 6 hours). This change went hand-in-hand with a further division of the work into a deterministic

|6

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

approach: KEMA, focusing on typical situations in a period of a year and covering both new time spans, and a stochastic approach (Delft University of Technology) addressing all 365x24x4=35040 PTU’s in a year but covering variations over ranges bigger than 1 PTU. The results of workpackage 2 are summarized in chapter 3 (paragraphs 3.5 and 3.6). In Annex 3-6 of the Annex Report the extensive research results of the substudies carried out under this workpackage can be found.

2.3

Workpackage 3: Instruments for balance control

In WP 3 the effectiveness of mitigating measures, such as fast start-up units, energy storage, short-term prediction of wind power and wind farm control strategies is assessed. Also international experience in this area is reported. The current and expected future resources (period 2008-2020) are discussed in greater detail in the following studies and activities: a. Fast start-up generation units; For the flexibility of the supply side related to the growth of wind energy (growth variability in supply, see WP 2) the possibilities of fast start-up units are explored. b. Wind power storage technologies; This inventory study results in a comparison of the effectiveness of various storage technologies c. Day-ahead wind power forecasts; This study provides insight into the accuracy of current forecasting methods and the opportunities for improvements. Also, the effect of minimizing wind power forecast errors by aggregation of geographically spread wind power is considered. d. Optimizing park layout; The breakdown of the project defined that a short study would be performed regarding the question to which extent the lay-out of wind parks may be optimized with respect to wind power fluctuation reduction. In 2006 the projectmembers decided, that this study wouldn’t be executed, because it doesn’t fit within the scope of the project. e. Wind farm control strategies (including cost impacts); An overview is presented of the results of work on wind farm control strategies. In addition, market bidding strategies that allow a wind producer to minimize its imbalance charges are explored. f. International experiences. This study compares different market designs with high wind penetration, such as Denmark, Spain and regions in Germany. This study generates ideas for a possible

|7

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

alternative approach to maintain balance and to assess application compatibility under Dutch conditions. The results of workpackage 3 are summarized in chapter 4. In Annex 4, 7-13 of the Annex Report the extensive research results of the studies carried out under this workpackage can be found. 2.4

Workpackage 4: Analysis and solutions

In the proposal of Si&Bh it was defined that in workpackage 4 recommendations and arguments would be carried out to be used in a discussion on the design of an imbalance market in the Netherlands which deals with large-scale wind power. This research has not been executed, because maintaining balance control with high wind power generation has a cross country character and cannot be solved only within the Netherlands. In chapter 5 we will close the chapter with an overview about the international dissemination of the project results.

|8

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3

Summary results workpackages 1 and 2 ‘Scenarios and analysis of current methods for balancing’

W.W. de Boer (KEMA), A.J. Brand (ECN), M. Gibescu (TU Delft) 3.1

Abstract

Wind power at sea is widely seen as the main contributor to a sustainable energy supply in the next decades. However the electricity market in the Netherlands was not a-priori designed for large penetrations of wind energy and may not be suitable. In the framework of Workpackages 1 and 2 of the We@Sea Project "System Integration and Balance Control with Large-scale Offshore Wind Energy", carried out by TenneT, Ecofys, KEMA, ECN and Delft University of Technology, this document presents an analysis of the barriers with respect to system balancing in the Netherlands if a large amount of wind energy is to be integrated into the power system. First, we assess the growth of conventional production and its flexibility, and define scenarios for the growth of wind energy, both up to the year 2020. Next, we explain how the wind speed data for simulating wind power production at system level are derived. On the basis of the growth scenarios, we then address the balancing challenges at system level due to wind power forecasting errors and also due to variations in the time scales within and larger than 15 minutes, up to 6 hours. By using a deterministic approach, typical situations and required control actions of Program Responsible Parties and the System Operator are identified. In addition, by using a stochastic approach overall statistics for the system balancing requirements are identified. Finally, we compare these requirements to the capabilities of the power system, using the projected conventional generation park for 2020.

|9

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.2

Introduction

3.2.1 Overview of this executive summary 







 

Chapter 3.3 presents the scenario's for the electricity sector in the form of a description of the current plant other than wind power and the scenario's for the other plant in 2020 (WP 1a en 1b,KEMA), and scenario's for wind energy growth until 2020 (WP 1c, Delft University) (Annex 1 and 2); Chapter 3.4 presents the creation of time series of measured and forecasted wind speed in the future wind farm locations out of such series in locations where wind was measured. This activity, facilitating the work in WP2, is a joint effort of ECN and Delft University of Technology (Annex 3 and 2); Chapter 3.5 presents wind power variations within PTU's (WP 2a), between PTU's separated 15 minutes and longer (WP 2b and WP 2c), and the effect on control at system level (WP 2d). This work is the outcome of the deterministic approach by KEMA (Annex 4); Chapter 3.6, presents wind power variations between PTU's separated 15 minutes, 30 minutes, 1 hour and 6 hours (WP 2b and WP 2c); the outcome of the stochastic approach by Delft University of Technology (Annex 6); Chapter 3.6.3 , presents a statistical analysis of wind predictability, as assessed jointly by ECN and Delft University of Technology (Annex 6); Chapter 3.7 presents insight into the flexibility and the variability of electricity production on the basis of different scenarios for the development of (wind) power plants in the Netherlands until the year 2020 and an assessment of the impact of large-scale wind power on the flexibility of conventional plants (Expected Results 1, 2).

| 10

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Schematic overview chapter 3 3.2

Introduction

3.2.1

Overview of this executive summary

3.3

Scenario’s for the electricity sector

3.3.1

Description and growth of other plant

3.3.2

Flexibility of production

3.3.3

Growth of wind energy

3.4

Time series of measured and forecasted wind speed

3.4.1

Methodology

3.4.2

Time series of measured wind speed

3.4.2.1

Transformation to turbine hub height

3.4.2.2

Identification of daily patterns and interpolation to wind farm locations

3.4.2.3

Identification of random variations and interpolation to wind farm locations

3.4.2.3

Transformation to 15 minutes averaging time

3.4.3

Time series of forecasted wind speed

3.5

Variations at different scales and control response - Deterministic approach

3.5.1

Deterministic approach

3.5.2

Power variations within 1 PTU

3.5.2.1

Methodology

3.5.2.2

Power variations due to normal wind speed variations

3.5.2.3

Power variations due to strongly correlated “gusts”

3.5.3

Power imbalance over more than 1 PTU

3.6

Variability and predictability of wind power production – the Stochastic approach

3.6.1

Methodology

3.6.2

Statistics of Wind Variability at Different Time Scales

3.6.3

Statistics of Wind Predictability

3.7

Summary OF Required AND available flexibility of production

3.8

References WP 1&2

| 11

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.3

Scenarios for the electricity sector

3.3.1 Description and growth of other plant First the retirement of the conventional production installations in the Netherlands was assessed. We assumed that power plants can operate 150% in respect to the original design (current practice). This leads to a decrease of 21 GW in 2005 to 9 GW in 2020 (= –57%). It is expected that most of the coal fired power plants and gas-fired combined cycle plants (STAGs1) are still operating. Nevertheless quite a lot of new power has to be installed in the next decades. Subsequently several scenario’s for new production capacity were set up for the years 2010 – 2015 – 2020. The parameters were economic growth (respectively 1, 2 and 3% per year), fuel mix (basic scenario with current gas-to-coal ratio 1.0:3.5, a gasand-coal reign scenario) and intensity of wind energy. The results for the year 2020 are shown in Table 1. Table 1 Installed power for several growth scenarios in 20202 Production in the Netherlands for several scenarios 2005 Type of Power production

Gas Motor Gas Turbine STAG of Combi

KEMA data base

2020 basic scenario

gas scenario

coal scenario

high growth scenario

low growth scenario

MW

MW

MW

MW

MW

MW

1.450

1.950

1.950

1.950

2.260

1.680

890

1.200

1.200

1.200

1.390

1.040

11.690

17.470

18.920

15.570

19.950

15.310

Conventional: boiler + ST (gas)

2.100

360

360

360

360

360

Conventional: boiler + ST (coal)

4.180

5.630

4.180

7.530

6.510

4.850

Nuclear

450

450

450

450

450

450

Waste and biomass

390

520

520

520

610

450

Wind Total production

1.200

7.800

7.800

7.800

10.400

4.800

22.350

35.380

35.380

35.380

41.930

28.940

The major growth for the basic scenario will most probably be of the CHP3 plants. Newer power plants, build after the year 2000, will have better materials and better control capabilities (~8% of nominal power per minute for gas, ~3%/min for coal). The range of power change capability for CHP plants is 50% or more. The control capabilities of the other scenarios differ slightly. For the coal scenario, the rate of power change capabilities will decrease a bit and for the gas reign scenario it will slightly increase.

1 2 3

Steam and Gas turbine Values for wind energy are discussed in chapter 3 Combined Heat and Power | 12

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.3.2 Flexibility of production Flexibility of production is required in order: 1 To follow the expected wind power variations, and 2 To compensate unexpected wind power variations, requiring a certain margin and rate of change capability (primary for PRP and secondly for TenneT). The flexibility of production can be defined in: 1 Rate of change of the total capacity, 2 Amount of regulating (spinning) and reserve power, 3 Rate of change of the spinning reserve units, and 4 Start time of the remaining units that are not delivering power during the load following cycle. Most of these items depend on the operating point in the load following cycle and on the types of power units operating in that operating point. The primary intention of the study is to give a rough indication of the flexibility. This is addressed in chapter 6. We expect a maximal ramping capability of 8%Pnom/min for gas-fired units and 3% for coal-fired units. The morning shoulder covers in the year 2020 approximately 10 GW (= difference between off-peak and peak load) with a maximal required ramp rate of 60 MW/min in 2020. We expect that the gas fired power units will carry this ramping load. Minimal 10 GW of gas-fired units have to be spinning. If they have an average rate of change of 4%/min, then 400 MW/min can become available. This is enough to handle the expected variability due to load. At the end of this document we will consider whether this anticipated flexibility is sufficient to cover for the combined wind power and load variations. 3.3.2.1 Growth of wind energy We have defined 3 scenarios for our analysis: slow, basic and advanced growth of wind energy, see Table 2. Table 2 Slow Basic Advanced

Installed nominal onshore and offshore wind power for three growth scenarios 2010 2015 2020 on off shore totaal on off shore totaal on off shore 1500 720 2220 1300 2010 3310 1000 3800 1500 1180 2680 1600 3110 4710 1800 6000 3120 6110 1600 1520 2000 4110 2400 8000

totaal 4800 7800 10400

An aim of the Dutch government (from the 2004 policy) was to have 20% of demand served with help of renewable energy in 2020. The scenario advanced growth will cover this completely with wind energy (given capacity factors of 25% and 37% respectively for onshore and offshore). This however is a rather optimistic view of wind energy growth. We therefore employed the basic growth scenario in this study and implemented in such a way that the other scenarios could be derived from it.

| 13

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

The locations of wind farms for the basic growth scenario 20204 offshore are given in Figure 1. The offshore locations were derived from the requests for permits for wind farms in the North Sea as filed by early 2006.

Figure 1

Location of offshore wind farms (stars) in 2020 according to the basic growth

scenario. Circles are locations in the KNMI weather data measurement network (http://www.offshore-wind.nl/default.asp?DOC-ID=23602

4

Aim was decreased in halfway 2006, current values are listed at http://www.senternovem.nl/offshore%5Fwindenergie/

| 14

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4

Time series of measured and forecasted wind speed

3.4.1 Methodology Estimates of system balancing energy require time series of wind energy production and prediction. These time series must be simultaneous series in the locations where wind farms will be placed. The future locations are determined based on the scenarios presented in section 2.3. As to the individual time series correct temporal and spatial correlations are taken into account. Wind speeds were collected from measurement locations for one year (1st June 2004 to 31st May 2005) by ECN. ECN also made wind speed forecasts for these measurement locations. Delft University converted the data into wind speed measurements and forecasts for the future locations of the wind farms (25 offshore locations and 14 onshore locations in the year 2020). Thus for every wind farm location there is a wind speed measurement and forecast signal for a 1-year period with a resolution of 15 minutes5. The one year of measured wind speed time series in the future locations were created with the following steps, which are detailed in section 3.2 (see also references [5-6]):    

The measured wind speeds are transformed to turbine hub height; Daily patterns and random variations are identified; Using interpolation for the patterns and multivariate normal theory for the variations the wind speed time series in the future locations are determined; Subsequently these 10-minute time series are transformed to 15-minute ones.

The forecasted wind speed time series in the future locations were created from one year of day-ahead forecasts of the wind speed at hub height in 7 locations, also by using interpolation and multivariate theory. Details are presented in section 3.3.

5

These signals will also be used in Workpackage 3.

| 15

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4.2 Time series of measured wind speed 3.4.2.1 Transformation to turbine hub height Figure 2 presents the locations of wind speed measurements in the Netherlands and the Netherlands Exclusive Economic Zone. There are 6 offshore, 4 coastal and 6 onshore locations.

F3 Huibertgat K13

Lauwersoog

Texelhors

Leeuwarden

IJmuide

Marknesse

M. Noordwijk Europlatform Hoek v. Holland L.E. Goeree

Lelystad de Bilt

Gilze-Rijen

Vlissingen

Figure 2

Wind speed measurement locations (reference Annex 6, source KNMI)

Since the wind speeds originate from routine weather masts, they are measured at sensor height zs only - usually but not always 10 meter. We therefore had to transform the average wind speeds μu to turbine hub height zh of 80 meter, to which end the wind speed standard deviation σu, either measured or estimated (for offshore locations), is employed. For further details, see reference [3].

| 16

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4.2.2 Identification of daily patterns and interpolation to wind farm locations Over a long period the wind speed can be decomposed into an average and a variation about that average. Since the variation may depend on the average, we had to suppress any relation and for this reason considered the logarithm of the wind speed rather than the wind speed.

2.5

Log Wind Speed (m/s)

observed computed

2

1.5

1

Figure 3

0

50 100 Time (10 min. intervals)

150

Daily wind speed patterns at the measurement and interpolated locations (reference Annex 6)

The data shows (Figure 3) a location-dependent daily pattern with a maximum around noon for onshore sites and an almost flat profile for offshore sites. The daily pattern at a new location now is obtained by interpolating between nearby locations.

| 17

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4.2.3 Identification of random variations and interpolation to wind farm locations The data also shows (Figure 4) a covariance that essentially decays with the separation between the locations. We use this characteristic to model the covariance with an exponential decay law, which allows us to treat the variation as a multivariate normal process with mean and standard deviation following from the sample statistics. 0.45

Covariance log wind speed

0.4

0.35

0.3

0.25

0.2

0.15

0.1

Figure 4

0

0.5

1

1.5

2 2.5 Distance (m)

3

3.5

4 5

x 10

Wind speed covariance as a function of separation between locations (reference Annex 6)

| 18

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4.2.4 Transformation to 15 minutes averaging time Having determined the log wind speed in the new locations, the time interval is changed from the meteorological standard 10 minutes to the 15 minutes required in the electricity sector. To this end weighted averaging is employed. Eventually this gives the wind speed time series in the locations where wind farms will be located. Figure 5 gives an impression of the variation between the wind speeds in the sites during a week.

Measured Wind (m/s)

15 most correlated least correlated 10

5

0

0

100

200

300

400 500 600 700 Time (10 min. intervals)

800

900

1000

0

100

200

300

400 500 600 700 Time (10 min. intervals)

800

900

1000

Estimated Wind (m/s)

15

10

5

0

Figure 5

Wind speed variations for measured and estimated locations (reference Annex 6)

| 19

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.4.3 Time series of forecasted wind speed Figure 6 presents the locations where wind speed is forecasted. The forecasts are dayahead forecasts of the 15-minute averaged wind speed at turbine hub height and were made with the wind power forecasting method AVDE of ECN.

F3

FINO

K13

EWTW1 NSW

Europlatform

Figure 6

Cabauw

Wind speed forecast locations (reference Annex 6)

Daily patterns and random variations where identified in the wind speed forecast time series, and interpolation (for the patterns) and multivariate theory (for the variations) were used in order to create time series of forecasted wind speed in the locations where wind farms are located. Figure 7 shows the patterns in the forecasted wind speeds and Figure 8 shows examples of the original and interpolated forecast error data at a number of locations.

| 20

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

1.8 observed interpolated

1.6 1.4

Forecast Error (m/s)

1.2 1 0.8 0.6 0.4 0.2 0 -0.2

0

Estimated Forecast Error (m/s)

Actual Forecast Error (m/s)

Figure 7

10

20

30

40 50 60 Time (15 min. intervals)

70

80

90

Daily patterns for wind speed forecast errors (reference Annex 6)

10 most correlated least correlated 5

0

-5

0

100

200

300 400 Time (15 min. intervals)

500

600

0

100

200

300 400 Time (15 min. intervals)

500

600

10

5

0

-5

Figure 8

Wind speed forecast errors for measured and estimated locations (reference Annex 6)

| 21

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.5

Variations at different scales and control response - Deterministic approach

3.5.1 Deterministic approach Now we analyse how large the imbalance may be for the basic wind power growth scenario 2020, distinguishing between:  

Imbalance within 1 Program Time Unit (PTU), 3.5.2, and Imbalance over more than 1 PTU, 3.5.3

Imbalance within 1 PTU is the sole responsibility for the system operator TenneT whereas imbalance covering more than one 1 PTU initially is the responsibility for the Programme Responsible Parties (PRP's). The analysis is based on a deterministic approach that identifies typical situations and gives an impression of the behavior of the TenneT controller (Frequentie Vermogen Regeling FVR) in these situations.

3.5.2 Power variations within 1 PTU 3.5.2.1 Methodology Based on available wind speed data we had to find the variations within 1 PTU. The available data consist of the 15-minute averaged wind speed signal and the standard deviation of an uncorrelated filtered white noise. Thus the total power variation within 1 PTU in our scope depends on the differences between the 15-minutes averaged wind speed signal (strongly correlated “gusts”) and the “sum” of power variations due to the normally distributed uncorrelated wind speed variations. The analysis is carried out under the assumption that the wind power forecast is optimal (= no systematic forecast error). 3.5.2.2 Power variations due to normal wind speed variations In Figure 9, the power variations at system level are shown resulting from 6000 MW offshore and 1800 MW onshore wind power. The wind speed varies around 10 m/s so the regime is partial load. The standard deviation of the power variation is 50 MW6.

6

Standard deviation is 30 (18) MW for 3100 (1200) MW offshore | 22

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Variaties binnen 1 pte; basis 2020: 7800 MW 200

150

100

MW

50

0

-50

-100

-150

-200 11.6

Figure 9

11.62

11.64

11.66

11.68 11.7 11.72 dagen vanaf 1 juni

11.74

11.76

11.78

11.8

Imbalance within 1 PTU in the case of 6000 MW offshore and 1800 MW onshore installed wind

power; 0.01 day corresponds to 14.4 minutes(reference [2])

The actions of the TenneT controller upon this kind of imbalance / power variations were analyzed with help of a system model that includes load, production and tie-lines flow, and frequency and secondary control (FVR) actions. In Figure 10 an impression of the structure of the model is given. Legenda: AI: Actual Interchange SI: Scheduled Interchange

FVR + 

PR (Program Responsible)

f: frequency

SI

bid allocation

P: Power P_ref_PV: reference signal from PR PR: Program Responsibility Party

- E-program - realisation strategy E-program

P_ref_PV

- internal feedback

: control signal from TenneT

f

AI

(measurements)

(measurements)

- production (+ load)

P prod AI

NL-part of UCTEgrid

Total load - ”power polisher” from sum of E-production P load - superposition of disturbances

Remaining

UCTE-grid

+ constant load + partly f

variable: fnl

controlled production units (primary control and AI-realisations)

Power Balance (seconds scale)

Figure 10

Model structure for analyzing the FVR-actions (reference [2])

| 23

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

In Figure 11 the response of the output of the controller is given for the 2015 and the 2020 basic growth scenarios.

Figure 11

Behavior of the FVR of TenneT during normal wind power variation (partial load) with 4710

MW and 7800 MW of wind energy (reference [2])

For the 2015 as well as the 2020 growth scenario the control efforts of the TenneT controller are not significant. The noisy signals are strongly filtered and only the relatively low frequencies pass slightly. It can be concluded that these variations do not result in significant additional control actions of TenneT’s controller. Normal variations of the TenneT controller without great disturbances are between –50 and +50 MW. 3.5.2.3 Power variations due to strongly correlated “gusts” The power variations due to strongly correlated “gusts” can be detected from the differences between consecutive 15-minute averaged wind speed signals. We performed the simulation for the basic growth scenario 6000 MW in 2020 and in Figure 12 only show the power gradient distribution.

| 24

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Distribution dP/dt basic scenario 6.000 MW off shore 100 88%

,

80

%

60 40 20 % 0,007 0,066 0,417

0 -200

Figure 12

-100

% 6,1

5,0 0,343 0,040 0,008 0,002

0 100 dP/dt -class, MW/min

200

X = 176 Y = 0.0019

Distribution of power changes, basic scenario 6000 MW offshore wind energy in 2020 (source

[2])

As can be seen, the majority of the time the rate of change of wind power at system level is small: 88% of time the rate of change is not faster than ±18 MW/min, whereas 99% of time the rate of change is not faster then ±35 MW/min. 3.5.3 Power imbalance over more than 1 PTU In order to give an impression of the imbalance due to forecast errors, we run the wind farm simulation model first with the forecasted wind speeds and then with the “real” wind speeds. Both runs result in wind power series, the forecasts for the Energy-programme and the "real" ones for the realized power. The difference is the imbalance. In Figure 13 we show the character of the imbalance. Onbalans tg v voors p elfouten Off sh ore 6.000 MW, b as is s ce nario 2020 2000 d ag 169, s td (P ) = 1277 MW d ag 48, s td(P ) = 1073 MW d ag 32, s td(P ) = 650 MW 1000

MW

0

-1000

-2000

-3000

-4000

0

5

10

15

20

25

u re n

Figure 13

Example of imbalance due to forecast errors in partial load (source [2]) | 25

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

The imbalance occurring in three typical days is depicted, resulting from an average 24 hour ahead stochastic wind speed forecast error (day 84, standard deviation = 1.3 m/s), a small stochastic forecast error (day 32, std = 0.7 m/s) and a large stochastic forecast error (day 169, std = 2.2 m/s). The character of these types of imbalance shows that the imbalance can be significant and can last for long. For this situation, not only regulating power but also reserve power is required provided the imbalance is assessed at least some PTU’s ahead with help of better wind forecasting. Thus the impact of wind power on the balancing system is large if the wind power is not forecasted well. A good forecast is vital particularly in partial load situations (which occur ~260 days/year). With the model as given in Figure 14 the actions of the TenneT controller were examined. The imbalance of day 84 was the input signal (basic growth scenario 2020: 6000 MW offshore and 1600 MW onshore).

FVR acties op onbalans tgv voorspelfouten van dag 84, std(v) = 1,3 m/s 4000 Som delta FVR Onbalans dag 84 (inverse) ACE 3000

MW

2000

1000

0

-1000

-2000

0

5

10

15

20

25

uren

Figure 14

Reaction of the FVR during day 84 due to a stochastic wind speed forecast error of 1.3 m/;

case 6000 MW offshore (source [2]

We show in Figure 14 the inverse imbalance (opposite sign) of day 84, ACE (Area Control Error) and the summed control signal (som delta FVR). We assume that TenneT reduces all imbalance due to the forecast errors. In the simulation the total regulating power was ±1500 MW7. If the imbalance reduction is optimal, then the inverse imbalance and the summed control signal are equal. As can be seen from the 7

Worst case, if the PRP’s are nor reacting. Practice in 2006: PRP’s take 20% of the total imbalance | 26

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

figure 8, this is the case when the absolute value of the imbalance is less than 800 MW. This simulation also gives an impression of the required rate of change for the control action. TenneT’s requirement of 7%/min of the offered regulating power, is able to handle these imbalances in this situation. Finally we mention that wind energy can result in scheduling problems in the case of high wind speeds at night. We have found that there is room for 2200 MW of wind power if there is import amounting to 3000 MW, and 4600 MW if import is not present. If more wind power is installed, additional shutting-down problems for the conventional production units will occur.

3.6

Variability and predictability of wind power production - Stochastic approach

3.6.1 Methodology In contrast with the detailed time-domain model with a one-minute resolution employed in paragraph 3.5, the stochastic approach in this chapter uses smoothed speed-power curves, which are simple, fast to compute, and appropriate for deriving the 15-minute average wind power production per location. For each of the projected locations, a wind-farm aggregated model has been created using regionally averaged speed-power curves, which depend on the area of the farm and the standard deviation of the wind speed distribution at the given farm's location. The multi-turbine curve is created by applying a Gaussian filter to a single-turbine speed-power curve. This results in a smoother output, particularly in the areas around the cut-in and cut-out speeds. For more details see reference [6]. A representative multi-turbine speed-power curve constructed for an offshore wind park of installed power 405 MW, at a location where the standard deviation of wind speed is 4.6 m/s, is shown in Figure 15.

| 27

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Wind Power (p.u.)

1.0

Single Turbine Offshore Park

0.5

0

0

5

Figure 15

10

15 Wind Speed (m/s)

20

25

30

Multi-turbine speed-power curve for a 405 MW offshore wind farm(reference Annex 6)

An example of the aggregation of real and forecasted wind power at system level for one week, resulting from the multi-turbine model applied to 15-minute averaged wind speed data, is shown in Figure 16.

6000 realization forecast 5000

Wind Power (MW)

4000

3000

2000

1000

0

0

Figure 16

100

200

300 400 Time (15 min. intervals)

500

600

Real and Forecasted Wind Power Profile for One Week, 6000+1800 MW Wind (reference

Annex 6)

| 28

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.6.2 Statistics of Wind Variability at Different Time Scales Table 3 presents the 99.7% confidence intervals and the extreme values (smallest and largest) of the 15 minute, 30 minute, 1 hour and 6 hour variations by the year 2020 for the Basic scenario (7800 MW). The sorted positive and negative variations in wind production over various time ranges are also shown in figure 17 for the same scenario. Based on the 99.7% confidence interval, the variations across 15-minute intervals are determined to be in the range of plus/minus 10% of the installed power for this scenario. Table 3

Statistics of Wind Variability, Basic 2020 Scenario (7800 MW)

Time Range

Min (MW)

Max (MW)

99.7% C.I. (MW)

15 min.

-1810

1918

[-794.8 .. 852.8]

30 min.

-2398

2795

[-1200.0 .. 1221.8]

1 hour

-2683

3669

[-1807.8 .. 1744.2]

6 hours

-6742

6327

[-4678.0 .. 4619.6]

Figure 17

Sorted variations in 7800 MW aggregated wind power (reference Annex 6)

Figure 17 presents the 99.7% confidence intervals for the wind variations across 15minute intervals for year 2010, 2015 and 2020 in the Basic scenario.

| 29

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

99.7% Confidence Interval for Wind Variations (MW)

1000 Negative 15 min. variations Positive 15 min. variations

800 600 400 200 0 -200 -400 -600 -800 2010

Figure 18

2015 Year

2020

Confidence intervals for 15-minute wind power variability, Basic Scenario (reference Annex 6)

3.6.3 Statistics of Wind Predictability Table 6 presents the 99.7% confidence interval plus the average, standard deviation, minimum and maximum of the imbalance (equal to the wind power forecast error) by 2020. Statistics for the positive and negative imbalance are presented in rows 2 and 3 of the same table. Based on the 99.7% confidence interval, in the scenario with 7800 MW of wind power by 2020, the positive (up regulation or reserve) balancing energy requirement is about 50% and the negative (down regulation) requirement is about 45% of the installed capacity. Statistics of Wind Predictability (MW), Basic 2020 Scenario (7800 MW)

Imbalance

Min

Max

99.7% C.I.

Mean

Stdev.

-5148

5326

[-3907.4 .. 3481.9]

40.7

1013.2

Positive

0

5326

[1.1 .. 3900.1]

721.8

693.6

Negative

-5148

0

[-4172.3 .. -1.0]

-724.5

729.8

Total

In Figure 19 we present the probability density function for the aggregated forecast error corresponding to the Basic 2020 scenario. By contrasting the empirical histogram (grey) with the fit to a normal distribution of the same mean and standard deviation (red curve), we show that the forecast errors in our data set are actually more concentrated around the mean, and have tails that decay faster than the normal, with the exception of a few isolated outliers.

| 30

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

-3

1

x 10

Forecast Error: 6000+1800 MW Fit to Normal Distribution

Probability Density

0.8 Sigma: 1013.16 0.6

Mu:

40.73

0.4

0.2

0 -6000

Figure 19

-4000

-2000 0 2000 4000 Aggregated Forecast Error (MW)

6000

Aggregated power forecast error probability density function for 7800 MW Wind (reference

Annex 6)

Figure 20 shows the 99.7% confidence intervals and standard deviations of the forecast-related imbalance for years 2010, 2015 and 2020 of the Basic scenario.

Standard Deviation and 99.7% Confidence Interval (MW)

4000

3000

Forecast Overestimate Standard Deviation Forecast Underestimate

2000

1000

0

-1000

-2000

-3000

-4000 2010

Figure 20

2015 Year

2020

Confidence intervals and standard deviations for balancing energy required due to forecast

errors, Basic Scenario, years 2010-2015-2020 (reference Annex 6)

| 31

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

In addition, we perform a statistical analysis of the MW forecast error data by prediction lag, which in our case varies between 48 and 144 Program Time Units (i.e. 15-minute intervals). The results are shown in Figure 21, expressed in percentage with respect to the installed power. As expected, the performance of the forecast degrades with prediction lag. The best values obtained are for the 12-hour (48 PTU’s) ahead forecast, where the standard deviation is 11.4%, and the 99.7% confidence interval is [-40.5% .. +40.0%].

Standard Deviation and 99.7% C.I. (%Pnominal )

0.8 0.6 0.4 0.2 standard devtn. overestimate underestimate linear underest. linear overest.

0 -0.2 -0.4 -0.6 -0.8 40

Figure 21

60

80 100 120 Forecast Lag (No. of 15-min. Intervals)

140

160

Standard deviation and 99.7% confidence interval by prediction lag (reference Annex 6)

Finally, table Table 4 presents two overall forecast error measures: the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE). The values are presented both in MW and in percentage with respect to the average power produced over the year, and with respect to the nominal (installed) power.

Table 4 RMSE MAE

Overall Predictability Statistics, 7800 MW Wind Power

(MW)

(%)Paverage

(%)Pinstalled

1014.0

29.5

12.9

723.1

21.0

9.2

| 32

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.7

Summary of required and available flexibility of production

Flexibility of production is required to compensate unexpected wind power variations due to forecast errors. The flexibility of production, needed to follow the expected wind power variations, will most probably be sufficient. The 7.800 MW installed wind energy in the 2020 Basic scenario will require an upward ramping ability from conventional generation of no more than 121 MW/min and a downward ramping ability of no more than 128 MW/min (see extreme values in the first row of table 4, chapter 5). Given the expected maximal required ramp rate of 60 MW/min during the morning shoulder in 2020 the simultaneous required ramp rate could be (in a worst-case scenario) 190 MW/min. This is less than the 400 MW/min ramp rate estimated to be available from peaking gas units in section 2.2. The variations within 1 PTU due to turbulence do not result in significant additional control actions for the controller of TenneT (FVR) (section 4.2.2), in case the wind energy produced during the PTU is adequately forecasted. However, wind energy can result in scheduling problems in the case of high wind speeds at night. Base load units have to be shut down which is not economic. Solutions have to be found to handle this phenomenon. Due to the nature of wind, forecast is never perfect, which causes unexpected power imbalances if no counter-actions are taken after the scheduling based on the forecast 24 hours ahead. We have made a statistical analysis of these imbalances based on the entire one-year data of real wind speed measurements and forecast values (chapter 5). The imbalance power due to forecast errors 12-36 hours ahead varies between +3480 MW and -3900 MW for the 7.800 MW installed wind power in the 2020 Basic scenario, given a 99.7% confidence interval. This means that the extremes will not be exceeded for more than 1.1 days/year. This requires between 45% down-regulation and 50% up-regulation or reserve power (percentages with respect to the 7800 MW installed capacity). The up-regulation amount is about 5.5 times bigger than the current N-1 reserve requirement (~ 700 MW), set for the largest conventional unit outage in the Netherlands. It can be concluded that the impact of large amounts of wind energy needs additional reserve requirements to deal with additional variability at the system level. In a second part of the study (WP 3 & 4) alternative methods will be analyzed to solve these problems by means of: better, short-term forecasting, wind farm control, additional fast power units and/or storage systems.

| 33

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

3.8

References WP 1&2

1

Systeemintegratie en balanshandhaving bij grootschalige windenergie op zee, Beschrijving onderzoeksvoorstel, 2004.

2

W.W. de Boer, W. van der Veen, N. Moldovan, Balanshandhaving met 6.000 MW aan windenergie. Deel 1: Balanshandhaving met huidige middelen (WP1 en WP2), maart 2006, 30450054-KEMA Consulting 2006-0123.

3

A.J. Brand, Observed and predicted wind speed time series in the Netherlands and the North Sea - Workpackage 2 We@Sea project System Integration and Balance Control, ECN, Report ECN-C--06-007, 2006.

4

M. Gibescu, B.C. Ummels, W.L. Kling, “ Statistical Wind Speed Interpolation for Simulating Aggregated Wind Energy Production under System Studies”, 9th International Conference on Probabilistic Methods Applied to Power Systems (PMAPS), Stockholm, Sweden, 2006.

5

M. Gibescu and A.J. Brand, “Estimation of System Balancing Requirements due to the Integration of Large-Scale Wind Energy”, 6th International Workshop on the large-scale integration of offshore wind energy, Delft, 2006.

6

M. Gibescu, A.J. Brand and W.L. Kling, “Estimation of Variability and Predictability of Large-scale Wind Energy in The Netherlands:, WIND ENERGY Wind Energ. Published online in Wiley Interscience (www.interscience.wiley.com) DOI: 10.1002/we.291, 2008.

| 34

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

4

Summary of Workpackage 3 ‘Instruments for Balance Control’

W.W. de Boer, A.J. Brand, M. Gibescu, B.C. Ummels 4.1

Introduction

In the framework of Workpackage 3 of the We@Sea Project "System Integration and Balance Control with Large-scale Offshore Wind Energy", carried out by TenneT, Ecofys, KEMA, ECN and Delft University of Technology, this summary presents an overview of the effectiveness – given wind energy growth scenarios for the Netherlands created during Workpackage 2 – of the following means for balance control: short-term wind speed forecasts, storage and fast start-up units, and regulation strategies. In the workpackage it was defined that ECN will set up a short study which examines how the park layout can be significantly optimized to generate the most constant output. Because this study doesn’t fit within the scope of the project, it wasn’t executed. In addition, KEMA has prepared a survey of international experience with the above topics. 4.2

Overview of this summary

This summary presents the results of WP3. In chapter 4.3, the possibilities of fast start-up units and energy storage are explored (TU Delft, Ecofys and KEMA). In chapter 4.4, day-ahead wind power forecasts are discussed (Ecofys, ECN, TU Delft). Chapter 4.5 presents the result of work on wind farm control strategies and the market bidding strategies coinciding these (ECN, Ecofys, KEMA, TU Delft). Finally, international experiences with power balancing means are discussed in chapter 4.6 (Ecofys, KEMA). Overview chapter 4

4.3

Fast start-up units and Compressed Air Energy Storage (KEMA, TU Delft, Ecofys)

4.3.1

Dedicated capacity for balancing wind power

4.3.2

Intelligent charging/discharging opportunities

4.3.3 4.4

Opportunities for Compressed Air Energy Storage (CAES) Day-ahead wind power forecasts (Ecofys, ECN TU Delft)

4.4.1

Wind power variability and prediction errors

4.4.2

Influence of Forecast Lag on the Required Regulating Energy

4.4.3

Aggregation of forecast errors and power imbalances

4.5

Wind Power Strategies (ECN, Ecofys, KEMA, TU Delft)

4.5.1

Control strategies for Dealing with Large Power Variations

4.5.2

Bidding Down-Regulating Strategies

4.6

International Experiences (Ecofys, KEMA)

4.6.1

Balancing markets

4.6.2

Literature survey

| 35

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

4.3

Fast start-up units and Compressed Air Energy Storage (KEMA, TU Delft, Ecofys)

The variability of large-scale wind power requires additional flexibility from the rest of the generation system. The combination of wind power generation with energy storage plants is frequently regarded as an optimal combination to deal with the variability of wind power. Also, fast start-up units may provide interesting opportunities. In this chapter, different energy storage technologies and fast start-up units are investigated and conclusions are drawn regarding possible synergies with wind power. 4.3.1 Dedicated capacity for balancing wind power First, the effectiveness of various storage technologies and fast-start up units are assessed on a back-to-back basis with wind power. This is done using the forecast power imbalances of wind power obtained in WP2. The investigated technologies are a 3000 MW, 10 GWh pumped storage system in combination with 852 MW of fast-start up units (open cycle gas turbine, OCGT). As can be seen from Figure 22, the use of energy storage and fast start-up units decreases the overall power variations: in case wind power falls short of the prediction, generation of energy storage and OCGT can be used, in case wind power exceeds the predicted value, energy storage is used to take in the additional wind power. Clearly, the impacts of wind power forecast errors are reduced by this configuration. When balancing wind power, it must be taken into account negative power imbalances (wind power falls short) has different impacts on the system than positive power imbalances. In case wind power falls short, additional generation is needed and units with fast ramp-up capabilities or short start-up times are required. In case wind power is long, generation levels of units in operation will need to be reduced, requiring flexibility with respect to ramp-down or short shut-down capabilities. In the report, negative power imbalances have been considered in more detail for a number of combinations of energy storage and fast start-up unit capacities. The larger the reservoir of energy storage, the larger the time energy storage will be able to back-up negative wind power forecast deviations. It is found that in order to arrive at a system reliability comparable to present levels at least a installed capacity of 2.6 GW of fast start-up units is required to balancing wind power shortfalls, as well as a 10 GWh reservoir for energy storage. On average, these units would be operating only 23% of the time. It should be noted that these number partly result from the fact that wind power is balanced by dedicated capacity; would a more system-wide approach be applied (wind power imbalance as part of the total system imbalance), different results would be obtained.

| 36

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Forecast Imbalance (MW)

4000 Original

2000

Reduced

0 -2000 -4000 1

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.4

1.45

1.5 4

10000

5000

0 Fast-Startup Power (MW)

Storage Energy (MWh)

x 10

1

1.05

1.1

1.15

1.2

1.25 1.3 PTU Number

1.35

1.25 1.3 PTU Number

1.35

1.4

1.45

1.5 4

x 10

800 600 400 200 0

Figure 22

1

1.05

1.1

1.15

1.2

1.4

1.45

1.5 4

x 10

Overall power deviations due to forecast errors, energy storage reservoir level and operating

points of OCGT for a simulated time-series (reference Annex 7)

4.3.2 Intelligent charging/discharging strategies In this section, two installed pumped storage systems (2 GW and 3 GW installed capacity, reservoir size 30 GWh) are compared for an intelligent control strategy. This strategy always aims at returning the storage system to a half-charged state (15 GWh), thereby allowing maximum flexibility with respect to the wind power forecast (energy storage may be used both for generation and storage at all times). Furthermore, a system-wide approach mentioned above is applied: the pumped storage is operated on a market-basis, making profits from the differences in peak and off-peak prices. Obviously, the high level of wind power examined in this research can be expected to have an impact on power prices on the market: high winds leads to low prices and vice versa.

| 37

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

4.3.3 Opportunities for Compressed Air Energy Storage (CAES) From the analyses performed in workpackage 3b (Annex 8), it can be concluded that the the technology for diabatic compressed air energy strorage (CAES) systems is available and has been applied successfully in a small number of cases. It is noted that many alternatives exist (conventional generation ramping, incidental wind power curtailment, extension of network capacity). Also, the profits of CAES will be difficult to assess, because these heavily depend on future power prices and the market design. It is concluded therefore that without dedicated support, CAES development in Europe will be restricted to a very limited number of initiatives (e.g. at already existing, unused salt domes). Whether such dedicated support may be justified by policy targets related to wind energy is still questionable.

4.4

Day-ahead wind power forecasts (Ecofys, ECN, TU Delft)

Improved wind power forecasting reduces wind power forecast errors and its associated power imbalances, and thereby facilitates a more economical integration of wind power into power systems. This paragraph provides insight into the accuracy of present forecasting methods and the opportunities of improvements. Also, the effect of minimizing wind power forecast errors by aggregation of geographically spread wind power is considered. 4.4.1 Wind power variability and prediction errors The value of wind energy for the Dutch electricity supply has been assess using one year of 15-minute wind speed time series for 7.8 GW wind power (Basic Growth 2020 Senario). It was found that the capacity factor was around 48%, with monthly means varying between 35% (summer) and over 50% (winter). Wind power production variations were found to vary between 0 an 2,2 GW/15 min, with variations between 1.5 and 2.2 GW/15 min. occurring twenty times per year, mostly due to simultaneous cut-out or cut-in because of storms or calms. Clearly, the spatial smoothing effect limits the power variations of the total installed wind power capacity. In order to investigate the impact of prediction age and imperfect predictions on wind power balance, the same method has been used to calculate the 15-minute predicted wind power for four different day-ahead forecasts (24h, 18h, 12h and 6h ahead). The accuracy of a prediction is evaluated by comparing the predicted values to the actual produced amounts. The forecast error of the power prediction is expressed by the Normalized Mean Absolute Error (NMAE). For the 24h ahead forecast, NMAE was found to have an average of 840 MW, while 70% of the forecast errors are within a 730 MW range, or 10% of installed capacity.

| 38

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

4.4.2 Influence of Forecast Lag on the Required Regulating Energy By analysing the four different day-ahead forecasts, it can be shown that the impacts of bad predictions can be avoided by making use of forecast updates, reducing the average prediction error to 720 MW using the 6 hour forecast. This clearly shows the importance of continuous wind power forecast updates, which also allow a better allocation of the forecast errors within the operation of other generation units in the system. It is shown that the forecast errors normalized standard deviation drops to half from the 36 to the 2-3 hours ahead prediction. Interestingly, the analysis also shows that predicted wind power variations are smaller than the actually realised wind power variations, which is especially for large wind power variations. Obviously, the latter conclusion depends on many factors and may be different for other wind speed data years. Furthermore, it should be noted that NMAE of NWP-based forecasts does not reduce to zero when the forecast horizon is 6 hours due to the intrinsic uncertainty in NWP models. Such a reduction however can be achieved if online production data is included in the forecasts, as is done in Figure 23, showing an example of the normalized standard deviation of the prediction error for the 0-36 hours ahead wind power forecast for 8000 MW capacity.

Figure 23

Forecast error standard deviation for 8000 MW wind power and forecast horizon (reference

Annex 8)

4.4.3 Aggregation of forecast errors and power imbalances This section specifies the impact of the aggregation of large-scale wind power on the overall forecast error and power imbalance due to wind power forecast errors. Two aggregation levels are considered: at the systems level and at the programme responsible party (PRP) level, being 7 individual market parties each with some wind power as part of their portfolio. The hypothesis was that a central aggregation of

| 39

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

power imbalances would allow some internal canceling out of wind power variations. It is found that this is indeed the case: aggregation at the system level requires less 5– 15% overall reserves for the balancing of wind power variations and the compensation of forecast errors (Table 1). Table 1

Power variations and forecast power variations for individual PRPs and Systems

15-min variation

forecast error

Max. Min. [MW/PTU] [MW/PTU]

99.7% C.I. [MW/PTU]

Max. [MW]

Min. [MW/PTU]

99.7% C.I. [MW]

Sum PRPs

+1915

-2294

[-1030 ... 969]

+5257

-5450

[-3718 ... 3972]

System

+1810

-1980

[-853 ... 799]

5148

5326

[-3482 ... 3907]

Difference

105

314

-177 ... 170

109

124

[-236 ... 65]

4.5

Wind Power Control Strategies (ECN, ECOFYS, KEMA, TU Delft)

Offshore wind farm control strategies may reduce the power variations of wind power. For all control measures it applies, that the wind power output can never exceed the available wind power. Thus, wind power may be reduced at all times, but for upward control actions, an initial decrease of wind farm output may be required. It can be noted that the design of the wind turbine, in particular the presence or absence of a cut-out wind speed, has an important impact on the opportunities for wind farm control. Concluding, it should be borne in mind that any wind farm control strategy brings about some opportunity loss of wind power production. The extent to which such strategies may be regarded as optimal therefore heavily depends on the market design for wind power, subsidy schemes and other, non-technical factors. 4.5.1 Control Strategies for Dealing with Large Power Variations A number of technical possibilities exist for controlling the wind park output, which is shown in Figure 24. The possibilities include an absolute output level limitation, ramping for balance control, ramp rate limitation (available only up to a certain extent for downward regulation) and delta control (fixed operation percentage of maximum available).

Figure 24

Wind farm control strategies applied at an existing offshore wind farm (reference Annex 4)

| 40

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

An estimate has been made of the benefits of control actions for the integration of large-scale wind power, using the wind power output time-series obtained in WP2. Furthermore, the predictability of large wind power variations is estimated with and without the application of these control strategies. The regulating strategies assessed here investigate the limiting of ramp rates and simultaneous cut-out in particular. The results show that ramp rate limitations indeed enable a reduction of the maximum wind power output variations. Also the possibility of ‘storm control’, in which a gradual shut down of the turbine is assumed as compared to a sudden turbine cut-out, allows a reduction of large wind power variations in number and in size. It is shown that the possibilities of both strategies depend on the maximum power variation level deemed acceptable. For limiting the down ramp, it may be necessary to reduce wind power output as a precaution as well: predicted variations may not happen, and variations may be different then predicted. 4.5.2 Bidding and Down-Regulating Strategies This section combines the wind power forecast and wind power prediction data (WP2) with price information from different markets. Price data have been obtained from the day-ahead spot market operated by the Amsterdam Power Exchange (APX) and from the hour-ahead single-buyer imbalance market operated by the Dutch TSO TenneT. Combining these data allows the computation of spot-market revenues and imbalance penalties or gains for large-scale wind power. Three different cases are evaluated:   

Perfect wind power forecast (zero prediction error); Median bid (realistic prediction error, forecast median bid into spot-market); Optimal quantile bid (realistic prediction error, forecast median bid into spotmarket).

The optimal quantile is determined from the probability distribution function of the random 15-minute average wind power production, and depends on price predictions for the APX day-ahead and TenneT imbalance prices. The wind producer may use two down-regulating strategies to avoid being charged for excess wind. The yearly revenues for various combinations of bidding and down-regulating strategies are then computed for an example wind producer. The results are shown in table 10.

| 41

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

Revenues for different bidding strategies

From the results, it can be concluded that bidding the optimal quantile combined with down-regulating to zero output results in the largest yearly revenues. It is interesting to note that for this data set, the optimal quantile bidding combined with a strategy of down-regulating to contract level produces a comparable income with that of an equivalent producer with a perfect wind power forecast.

4.6

International Experiences (Ecofys, KEMA)

In this paragraph, a comparative inventory is made of different market designs in countries with high wind power penetrations, such as Denmark, Germany and Spain. These market designs are then compared to the existing balancing market design in the Netherlands. Furthermore, some recent international papers on various topics discussed in WP3 are discussed. This work is done for this summary. We didn’t carry out an extensive study/report. A more comprehensive research of international experiences has been carried out under Annex 25 of the IEA Wind Implementing Agreement ‘Design and Operation of Power Systems with Large Amounts of Wind Power’. 4.6.1 Balancing markets The analysis of the approaches to balancing services has revealed major differences in the four countries. First, differences can be found in the institutional environment. The responsibility for balancing service markets can either be assigned to the system operator (Germany, Spain, Denmark (onshore wind power only)) or an associated market party (the Netherlands, Denmark (offshore wind power only)). Second, differences exist in the rules of the use and provision of balancing services. The following conclusions can be drawn: 

Regarding the short-term markets (intra-day) progress has been made in the past years to increase the liquidity of short term markets that are important for wind energy. Gate closure times of about 1 hour ahead of delivery are sufficient to include maximum forecast accuracy;

| 42

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE





Systems with dual imbalancing prices pose a number of problems for wind energy, which has led to huge losses in the case of UK. To minimise imbalance costs, producers or traders have to aggregate their generation portfolios; In all analysed systems, with the exception of Germany, operators or traders are responsible for wind power forecasting. Since the predictions are used to optimise financial results, a prediction has to be made by the system operator as well (except in case wind power balancing responsibility lies with the market, such as in the Netherlands and UK).

There is a clear European trend towards more cross-border balancing, which certainly promises advantages for wind power. Of course, wind power already is a transEuropean phenomenon from a power flow perspective. It has been shown that balancing geographically larger control areas provides benefits for wind power especially, not only because of decreased overall variability and unpredictability but also because of larger market volumes and increased balancing resources. Certainly, it is rather difficult to quantitatively assess the necessity or effectiveness the different technical solutions for balancing wind power. Furthermore, it should be noted that in all European countries investigated, a considerable portion of wind power revenues comes from national support schemes. On the other hand, at system level the benefits in terms of lower electricity process and reduction emission weigh against the support schemes. Controlling the power output of (offshore) wind farms must be considered as an option mostly from a power system operations point of view, since the opportunity loss by curtailment is significant. 4.6.2 Literature survey The literature survey was carried out in 2007. In the literature survey the projectpartners focussed on the Sixth International Workshop on Large Scale Integration of Wind Power and Transmission Networks for Off shore Windfarms, held in Delft in October 2007. They focussed on this workshop because, most of all relevant studies of market integration of wind power, offshore wind farms, power system balancing and costs and the integration and control of offshore wind farms were discussed at that workshop. Furthermore, a number of European markets has been investigated. In relation to the work of WP3, the following conclusions can be drawn: 



German experience shows that aggregation of wind power production over a larger geographical area smoothes out wind power variability and improves the quality of the wind power forecast due to partly uncorrelated forecast errors. As a result of this, the total amounts of power reserves decreases, both the reserves held and the reserves actually applied. Balancing wind power across German control areas is shown to be more efficient; German experience also shows that the quality of wind power forecasts significantly increases as the forecast horizon decreases. Improved wind power forecasts are used to optimise the commitment and dispatch of other generation

| 43

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE







units in the system. Power reserves held for wind power are decreased and possible be offered into the market and a more efficient use is made of available ramping capabilities between different units; The regional clustering of wind parks as a virtual wind power plant may provide benefits for reactive power control and active power management. The technical possibility of wind power plant control for power balancing is however generally not economically attractive in the present German market design. However, the temporal use of wind power for negative operating reserve may be attractive during certain periods; Energy storage should not be considered as a dedicated balancing technology for wind power. The benefits of dedicated energy storage systems are considerably lower than a more system-wide, market-based operation based on power prices. Energy storage may be used to minimise power imbalances, for primary control (frequency control), for transmission congestion relief and for power quality control. It is shown that large reservoir sizes relative to the generating/storage capacity are most attractive, but also costly, and therefore a multi-purpose operation strategy will be required; The use of intra-day trade and short-term wind power forecasts may reduce wind power imbalances.

| 44

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE

5

Workpackage 4: Analysis and solutions

In the proposal of Si&Bh it was defined that in workpackage 4 recommendations and arguments would be carried out to be used in a discussion on the design of an imbalance market in the Netherlands which deals with large-scale wind power. This research hasn’t been excecuted, because maintaining balance control with high wind power generation has a cross-country character and can’t be solved only within the Netherlands. Instead of carry out the tasks of workpackage 4 the results of the project (workpackages 1-3) are used by the projectmembers as an input for international papers and conferences. In paragraph 5.1 we give an overview about the international communication of the project results. 5.1

International communication about the project results

The results of the different workpackages were frequently communicated by the project partners, both in international journals as at international conferences, of which a number is presented in the Annex of this report (Annex 7, 12 and 13. The results are also used for other studies and research work. The most direct result is the Ph.D. thesis study of B.C. Ummels at the TU Delft8. His study is titled: “Power System Operation with Large-Scale Wind Power in Liberalised Environments”. The thesis is about the consequences of the integration of a lot of wind power for the existing power system. Questions that B.C. Ummels answers are: What problems do we run into and what solutions are available? Is it possible to produce one third of the electricity demand with onshore and offshore wind energy?

8

Ph.D. Thesis B.C. Ummels, “Power System Operation with Large-Scale Wind Power in Liberalised Environments”, TU Delft, 2009

| 45

A SUST AIN ABLE ENERGY SUPPLY FOR EVERYONE