Wind generation penetration into electricity markets

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

130

�

Abstract - The Vietnamese government have plan to develop the wind farms with the expected capacity of 6 GW by 2030. With the high penetration of wind power into power system, wind power forecasting is essentially needed for a power generation balancing in power system operation and electricity market. However, such a tool is currently not available in Vietnamese wind farms as well as electricity market. Therefore, a short-term wind power forecasting tool for 24 hours has been created to fill in this gap, using artificial neural network technique. The neural network has been trained with past data recorded from 2015 to 2017 at Tuy Phong wind farm in Binh Thuan province of Viet Nam. It has been tested for wind power prediction with the input data from hourly weather forecast for the same wind farm. The tool can be used for short-term wind power forecasting in Vietnamese power system in a foreseeable future.

Keywords: power system; wind farm; wind power forecasting; neural network; electricity market.

I. NECESITY OF WIND POWER FORECASTING

Today, the integration of wind power into the existing grid is a big issue in power system operation. For the system operators, power generation curve of wind turbines is a necessary information in the power sources balancing. From the dispatchers’ point of view, wind power forecast errors will impact the system net imbalances when the share of wind power increases, and more accurate forecasts mean less regulating capacity will be activated from the real time electricity market [1]. In the deregulated market, day-ahead electricity spot prices are also affected by day-ahead wind power forecasting [2]. Wind power forecasting is also essential in reducing the power curtailment, supporting the ancillary service. However, due to uncertainty of wind speed and weather factors, the wind power is not easy to predict.

In recent years, many wind power forecasting methods have been proposed. In [3], a review of different approaches for short-term wind power forecasting has been introduced, including statistical and physical methods with different models such as WPMS, WPPT, Prediktor, Zephyr, WPFS, ANEMOS, ARMINES, Ewind, Sipreolico. In [4], [5], the methods, models of wind power forecasting and its impact on

*Research supported by Gesellschaft fuer Internationale

Zusammenarbeit GmbH (GIZ). D. T. Viet is with the University of Danang, Vietnam (email:

[email protected]). V. V. Phuong is with the University of Danang, Vietnam (email:

[email protected]). D. M. Quan is with the University of Danang, Vietnam (email:

[email protected]). A. Kies is with the Frankfurt Institute for Advanced Studies, Germany

(email: kies@fias uni-frankfurt.de). B. U. Schyska is with the Carl von Ossietzky Universität Oldenburg,

Germany (email: [email protected]). Y. K. Wu is with the National Chung-Cheng University, Taiwan (email:

[email protected]).

the electricity market and power systems have been presented. The practice and experience of short-term wind power forecasting accuracy and uncertainty in Finland has also been investigated [1].

In general, the equation for wind power P (W) of each wind turbine is given by the formula (1):

P = (1/2)ρ×A×Cp×Ng×Nb×V 3 (1)

where ρ: air density (kg/m3), A: rotor swept area (m2), Cp: coefficient of performance, V: wind speed (m/s), Ng: generator efficiency, Nb: gear box bearing efficiency [6].

Unfortunately, many multiplication factors in the formula (1) are uncertain. It leads to uncertainty in relationship between wind speed and wind power of each wind turbine [7].

II. WIND POWER FORECASTING IN VIETNAM

A. Wind power in Vietnam

Vietnam is considered to have high potential for wind energy. The wind energy potential of Vietnam is shown in Table 1, Fig. 1 and Fig. 2 [8]:

TABLE 1. WIND ENERGY POTENTIAL OF VIET NAM AT 80 M ABOVE GROUND LEVEL

Average wind speed (m/s)

<4 4-5 5-6 6-7 7-8 8-9 >9

Area (km2)

95,916 70,868 40,473 2,435 220 20 1

Area (%)

45.7 33.8 19.3 1.2 0.1 0.01 0

Potential (MW)

956,161 708,678 404,732 24,351 2,202 200 10

The development of wind power has been paid attention by both the Vietnamese government and investors. The national renewable energy development strategy by 2030, which was approved by the Vietnamese Prime Minister, emphasizes the role of wind power in particular. Expectations about installed wind power capacity are 800 MW in 2020; 2,000 MW in 2025 and around 6,000 MW by 2030 [9], [10].

By 2017, 160 MW of wind power capacity has been installed, some large wind farms with capacity and year of beginning operation are listed - Tuy Phong: 30 MW (2009); Bac Lieu: 16 MW (2013), and 99.2 MW (2016); Phu Lac: 24 MW (2016); Phu Quy: 6 MW (2013) [8].

A Short-Term Wind Power Forecasting Tool for Vietnamese Wind Farms and Electricity Market*

Dinh Thanh Viet, Vo Van Phuong, Minh Quan Duong, Alexander Kies, Bruno U. Schyska and Yuan Kang Wu

978-1-5386-5126-1/18/$31.00 ©2018 IEEE

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

131

Figure 1. Wind resource map of Vietnam at the height of 80 m

Figure 2. Representative wind profile for the three regions in Vietnam

B. Wind power forecasting in Vietnam

At present, there is no effective tool for predicting wind power in Vietnam. With the increasing integration of wind energy into the Vietnamese power system, the projected capacity of wind power plays an important role in supporting the optimal operation of wind power plants as well as the electricity market.

The forecast error of the whole wind farm will be much affected by the forecast errors from all wind turbines as a sum. For the electricity market operators, the predicted power

of the whole wind farm at the point of coupling into the power grid is needed, rather than the sum of predicted powers of all turbines. In this paper, the approach of wind power forecasting for the whole wind farm will be investigated.

III. SHORT-TERM WIND POWER FORECASTING USING NEURAL NETWORK

A. Neural network

A neural network is a multi-input, multi-output system, consisting of an input layer, one or two hidden layers and an output layer. Each class uses a number of neurons, and each neuron in a layer is connected to neurons in the adjacent layers with different weights. The architecture of the typical neural network is shown in Fig. 3 [11], [18].

Figure 3. Neural network structure

where input X = (x1, x2, ..., xd), output O = (o1, o2, ..., on). The signal is fed into the input layer, passing through the hidden layer and to the output layer. In a neural network, each neuron (except neuron at the input layer) receives and processes stimuli (inputs) from other neurons. Each input is first multiplied by the corresponding weight, then the resulting products are added to produce a weighted sum, which is passed through a neuron activation function to produce the output of the neuron [11], [12].

B. Feedforward neural network

A feedforward neural network usually has one or more hidden layers of sigmoid neurons followed by a linear neurons output layer. The paper uses the model of a feedforward neural network as described in Fig. 4. The input layer consists of 3 neurons of historical wind speed, temperature and wind power. The neural network has 20 neurons in the hidden layer and 01 neuron in the output layer.

The neural network may be used as a general function approximator. With enough neurons in the hidden layer, any function with a specific number of discontinuities arbitrarily can be approximated well. The algorithm for building the neural network - wind power forecasting model is shown in

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

132

Fig. 5. The historical data from wind farm, including wind power, wind speed and temperature, are loaded into the program and stored as a matrix of forecasting variable. Because of different operation scenarios, the historical data may not correctly reflect the relationship between wind power to wind speed and temperature. In few cases, historical data in practice may show negative values of wind power during generator starting time. Therefore, a data preprocessing is needed for reducing forecasting error. The historical data is used for training the forecasting neural network.

Figure 4. Feedforward neural network.

C. Wind power forecasting model

From the formula (1), it is obvious that generating power of each wind turbine depends largely on the wind speed. The temperature of environment is chosen as second input, affecting on the output power [13], [14], sum of output powers from the turbines serves as another input for neural network wind power forecasting model (Fig. 6). The input data is used to forecast the generating power of the wind farm. This model is designed for Vietnamese wind farms’ power forecasting (short name: VWPF).

IV. CASE STUDY: WIND POWER FORECASTING FOR TUY PHONG WIND FARM

A. Simulation data

Tuy Phong is the first large-scale wind farm in Vietnam with a total capacity of 30MW, including 20 turbines of Fuhrländer, each turbine has a height of 85m, a blade diameter of 77m, and capacity of 1.5MW. The research is based on real data on wind speed and wind power production at Tuy Phong wind farm for 3 years. Collected data from January 1, 2015 to December 31, 2017 was used for wind power forecasting.

The data in the forecasting model VWPF includes wind speed, environmental temperature and wind power, which are collected every hour, represented by 24 lines per day. Example of data on October 7, 2017 for wind power forecasting is shown in Table 2.

The data set is divided into two data subsets:

� Data subset 1 from 01/01/2015 to 30/10/2017 is used to train the neural network. This is a database with

data collected in 24,816 hours, almost 3 years, large enough for the neural network training purpose.

� Data subset 2 from 01/11/2017 to 31/12/2017 is used to compare the forecast results with the collected actual data for the error evaluation purpose.

Figure 5. Algorithm for building the neural network - wind power

forecasting model VWPF.

Figure 6. Wind power forecasting model VWPF

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

133

TABLE 2. EXAMPLE OF DATA ON OCTOBER 7, 2017

Date Hour Tempera- ture ( )

Wind speed (m/s)

Wind power (W)

07/10/2017 1 24 3.56 272,299

07/10/2017 2 24 3.51 268,449

07/10/2017 3 24 3.52 266,840

07/10/2017 4 25 3.93 596,442

07/10/2017 5 25 3.9 572,175

07/10/2017 6 27 4.05 767,869

07/10/2017 7 27 3.96 772,839

07/10/2017 8 32 5.54 2,067,089

07/10/2017 9 32 5.47 2,137,362

07/10/2017 10 32 5.35 2,071,529

07/10/2017 11 32 5.4 2,044,341

07/10/2017 12 32 5.28 2,237,072

07/10/2017 13 32 5.25 2,105,929

07/10/2017 14 32 5.34 2,143,538

07/10/2017 15 30 4.68 1,683,934

07/10/2017 16 29 4.04 767,723

07/10/2017 17 24 4.53 1,632,532

07/10/2017 18 24 5.29 2,165,775

07/10/2017 19 24 5.77 2,943,629

07/10/2017 20 24 5.89 2,947,676

07/10/2017 21 24 6.13 3,200,192

07/10/2017 22 24 5.98 3,200,290

07/10/2017 23 25 5.75 3,084,564

08/10/2017 0 25 5.73 3,065,696

B. Error evaluation

The error measures - Mean Absolute Percent Error (MAPE), Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) are used to evaluate the accuracy of the forecasting model [15], [16].

The MAPE represents the accuracy of the model as a percentage of the error, calculated according to formula (4):

(4)

where:

Preal: actual power output of the wind power plant,

Ppred: generation power according to forecasting model,

N: number of forecasting data.

The MAE shows the accuracy of the model in the same unit of measure as the predicted data. This index is used to evaluate the margin of error and is calculated according to the formula (5):

(5)

In order to evaluate MAE in percentage for comparison between different models, we can use Normalized Mean Absolute Error (NMAE) (6):

(6)

where Pinst is the wind farm installed capacity.

The RMSE is the standard deviation of the prediction errors (residuals). This is also a frequently used measure of the differences between values forecasted by a model and the values actually observed. The RMSE is calculated by the formula (7):

(7)

Similarly, we can use Normalized Root Mean Square Error (NRMSE) in percentage (8):

(8)

C. Result

By putting Tuy Phong wind farm data into the model and implementing neural network training, the forecasted model has been received and shown in Fig. 7, where the blue line represents the predicted wind power, the red line represents the actual generated wind power from the data subset 1. Fig. 7 shows a part of the snapshots for series of predicted time, the total number of snapshots is 23,856 hours (from 1/1/2015 to 20/9/2017).

The forecasting model VWPF is validated with data from any date in the data subset 2 for 92 days during period from 01/10/2017 to 31/12/2017. The real data of previous day is updated and included into the training database for the next day wind power forecast. As an example, the forecast on November 26, 2017 (the lowest wind speed was 8.39 m/s and the highest was 15.44 m/s with the steady change of wind speed throughout the day) is shown in Fig. 8. The error in the graph represents difference between real wind power and predicted wind power. The measures for the accuracy of the forecast result on November 26, 2017 are: MAPE=4.72%; NMAE=4.66%; NRMSE=5.66%.

Forecast results for December 17, 2017 with the speed changes from 6.15 m/s to 14.47 m/s are shown in Fig. 9. The measures for the accuracy of the forecast result on 17/12/2017 are: MAPE=5.44%; NMAE=4.74%; NRMSE=6.24%.

The wind power day-ahead forecast results in one week from 25/12/2017 to 31/12/2017 and relevant forecast error are shown in Fig.10. The forecast error or difference between the observed and the forecast wind power for one week from 25/12/2017 to 31/12/2017 is evaluated by the following values: MAPE=8.78%; NMAE=4.14%; NRMSE=4.58%

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

134

Based on the results of the forecasted wind power from VWPF, we have a practical range of errors as in Table 3:

TABLE 3. RANGE OF ERRORS FROM PRACTICAL TESTING WITH THE MODEL VWPF

No MAPE (%) NMAE (%) NRMSE (%)

1 8.05 5.52 7.28

2 4.72 4.66 5.66

3 5.13 5.13 6.20

4 10.06 5.98 7.32

5 7.03 5.59 7.70

6 6.07 4.07 4.09

7 5.44 4.74 6.24

8 9.64 5.95 8.00

9 5.54 5.94 7.70

Average 6.85 5.29 6.69

Figure 7. Wind power forecasting result after neural network training.

Figure 8. Forecasted versus actual wind power on 26/11/2017

Figure 9. Forecasted versus actual wind power on 17/12/2017

Summary of error range from Table 3: MAPE = 4.72- 10.06%, NMAE = 4.07-5.98%, NRMSE = 4.09-8.00%.

2018 4th International Conference on Green Technology and Sustainable Development (GTSD)

135

Figure 10. Forecasted versus actual wind power during one week from

25/12/2017 to 31/12/2017

TABLE 4. COMPARISON THE FORECAST ERROR INDICES OF THE PROPOSED MODEL (VWPF) WITH SOME OTHER MODELS

Forecasting Model Error indices

MAPE (%)

NMAE (%)

NRMSE (%)

Persistence 14.43 6.18 7.99 BPNN 14.35 5.98 7.53

RBFNN 12.73 5.94 7.40 ANFIS 14.92 6.24 8.03 NNPSO 11.51 5.35 6.59

WT+BPNN 12.19 5.77 7.18 WT+RBFNN 11.18 5.62 6.95 WT+ANFIS 12.58 5.86 7.67 WT+NNPSO 8.19 4.86 6.28

VWPF 6.85 5.29 6.69 Table 4 showed comparison between the forecast errors

of the proposed model with some other published models [17]. The error indices in different seasons [17] were recalculated as the average values. From Table 4, we find that the error indices between of the VWPF model is relatively smaller in comparison with most of the published wind power forecasting models. It proves that the VWPF model provides reliable forecasting results.

V. CONCLUSION In the paper, a model of the wind power forecasting

(VWPF) is developed for this need in Vietnam. The power system operators are usually interested in the forecasting of the whole wind farm’s power, which is generated into the power system, rather than forecast power of each wind turbine. It shows advantage and effectiveness of the developed model in power prediction for the whole wind farm, which is well appropriate for dispatcher working as well as electricity market operator. The neural network prediction model can be used for short-time wind power forecasting (hour-ahead, day-ahead, and week-ahead). The forecasting model has been applied for estimating the wind power output of the Tuy Phong wind power plant in Binh Thuan province, Vietnam. The predicted results were evaluated with the average forecast error indices MAPE=6.85%, NMAE=5.29%, NRMSE=6.69%. The

forecast error indices, showing the high accuracy of the model, are relatively smaller in comparison with most of similar research models (Table 4). Application of artificial intelligence technique at the connected point of the wind farm to the power grid proved effectiveness of this approach. This wind power forecasting tool can be applied not only for Tuy Phong wind farm, but also for the others in Vietnam.

ACKNOWLEDGMENT

This work is part of the R&D Project “Analysis of the Large Scale Integration of Renewable Power into the Future Vietnamese Power System”, financed by Gesellschaft fuer Internationale Zusammenarbeit GmbH (GIZ, 2016-2018).

REFERENCES

[1] H. Holttinnen, J. Miettinen, S. Sillanpää “Wind power forecasting accuracy and uncertainty in Finland”, VTT Technology 95-320, 2013.

[2] T. Jónsson,; P. Pinson, H. Madsen, “On the market impact of wind energy forecasts”, Energy Economics, Volume 32, Issue 2, March 2010, pp. 313-320.

[3] X. Wang, P. Guo, X. Huang, “A Review of Wind Power Forecasting Models”, Energy Procedia, vol. 12, pp. 770 – 778, 2011.

[4] G. Kariniotakis, Renewable Energy Forecasting: From Models to Applications, 1st ed., Woodhead Publishing, 2011.

[5] G. Giebel, R. Brownsword, G. Kariniotakis, M. Denhard, C. Draxl, “The State of the Art in Short-Term Prediction of Wind Power: A Literature Overview”, 2nd ed., 2011.

[6] A. Sarkar, D. K. Behera, “Wind Turbine Blade Efficiency and Power Calculation with Electrical Analogy”, International Journal of Scientific and Research Publications, vol. 2, Issue 2, February 2012.

[7] D.M. Quan, E. Ogliari, F. Grimaccia, S. Leva, M. Mussetta, “Hybrid model for hourly forecast of photovoltaic and wind power”, 2013 IEEE International Conference on Fuzzy Systems, p.p 1-6, 2013.

[8] N. Q. Khanh, “Analysis of future generation capacity scenarios for Vietnam”, Green Innovation and Development Centre (GreenID), Vietnam, 2017.

[9] The Vietnamese Prime Minister, “Approving the development strategy of renewable energy of Vietnam by 2030 with a vision to 2050”, Decision No. 2068/QD-TTg dated November 25, 2015.

[10] A. Kies, B. Schyska, D. T. Viet, L. Bremen, D. Heinemann, S. Schramm, “Large-Scale Integration of Renewable Power Sources into the Vietnamese Power System”, Energy Procedia, vol. 125, pp. 207– 213, 2017.

[11] S. Samarasingh, Neural Networks for Applied Sciences and Engineering: From Fundamentals to Complex Pattern Recognition, CRC Oress, Taylor & Francis Group, Boca Raton, 2007.

[12] The MathWorks Inc, Neural Network Toolbox User’s Guide, 2014. [13] A. Singh ; K. Gurtej ; G. Jain ; F. Nayyar ; M. M. Tripathi, Short

term wind speed and power forecasting in Indian and UK wind power farms, 2016 IEEE 7th Power India International Conference (PIICON),

[14] Salih Mohammed Salih, Mohammed Qasim Taha, Mohammed K. Alawsaj, “Performance analysis of wind turbine systems under different parameters effect”, International Journal of Energy and Environment, vol. 3, Issue 6, pp.895-904, 2012.

[15] D. B. Alencar, C. M. Affonso, R. C. L. Oliveira, J. L. M. Rodríguez, J. C. Leite and J. C. R. Filho, “Different Models for Forecasting Wind Power Generation: Case Study”, Energies, 2017.

[16] X. Zhao, S. Wang, T. Li, “Review of Evaluation Criteria and Main Methods of Wind Power Forecasting”, Energy Procedia, vol. 12, pp. 761 – 769, 2011.

[17] P. Mandala, H. Zareipourb, W. D. Rosehart, Forecasting Aggregated Wind Power Pro duction of Multiple Wind Farms Using Hybrid Wavelet-PSO-NNs, International Journal of Energy Research, Vol.38, Issue13, pp. 1654-1666, 2014.

[18] S. Haykin, Neural Networks and Learning Machines, 3rd ed., Pearson Education Inc, 2009.

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012 809

Comparison of Wind Energy Support Policy and Electricity Market Design in Europe,

the United States, and Australia Néstor Aparicio, Member, IEEE, Iain MacGill, Member, IEEE, Juan Rivier Abbad, Member, IEEE, and

Hector Beltran

Abstract—This paper is intended to fill a gap in the current lit- erature comparing and contrasting the experience of a number of Europeancountries,U.S. states, andAustraliawithregard towind energysupportpolicyandelectricitymarketdesign.Aswindpene- trations increase, thenatureof thesearrangementsbecomesan in- creasingly importantdeterminantofhoweffectivelyandefficiently thisgeneration is integrated into theelectricity industry.Thejuris- dictions considered in this paper exhibit a range of wind support policy measures from feed-in tariffs to green certificates, and elec- tricity industry arrangements including vertically integrated utili- ties,bilateral tradingwithnetpools,aswellasgrosswholesalepool markets. We consider the challenges that various countries and states have faced as wind generation expanded and how they have responded. Findings include the limitations of traditional feed-in tariffsathigherwindpenetrationsbecausetheyshieldwindproject developersandoperators fromthe implicationsof theirgeneration on wider electricity market operation. With regard to market de- sign, wind forecasting and predispatch requirements are particu- larly important for forward markets, whereas the formal involve- ment of wind in scheduling and ancillary services (balancing and contingencies) is key for real-time markets.

Index Terms—Balancing markets, electricity market design, re- newable energy policy, wind energy.

I. INTRODUCTION

P OLICY measures to support greater wind energy havehadademonstrated impacton itsdevelopment indifferent jurisdictions around the world. Experience to date suggests that feed-in tariff (FIT) policies have been the most successful approach in rapidly expanding wind generation capacity, as demonstrated incountries includingDenmarkandSpain,which now have world leading wind energy penetrations [1]. This ap- proach, however, may cause increasing integration challenges for the electricity industry as wind penetrations continue to

Manuscript received August 30, 2011; revised June 08, 2012; accepted July 06, 2012. Date of publication September 10, 2012; date of current version September 14, 2012. This work was supported in part by the Universitat Jaume I under Grant P1·1A2008-11. N. Aparicio’s research visit to the Centre of Energy and Environmental Markets, which kindly offered him a visiting position, was supported by the Universitat Jaume I under Grant E-2008-06. N. Aparicio and H. Beltran are with the Area of Electrical Engineering, Uni-

versitat Jaume I, 12071 Castelló de la Plana, Spain (e-mail: [email protected]). I. MacGill is with the School of Electrical Engineering and Telecommuni-

cations and Centre for Energy and Environmental Markets, University of New South Wales, Sydney 2052, Australia (e-mail: [email protected]). J. Rivier Abbad is with Iberdrola Renovables, 28033 Madrid, Spain (e-mail:

[email protected]). Color versions of oneormore of thefigures in this paper are available online

at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TSTE.2012.2208771

rise. The value of electricity within a power system varies over time, by location and subject to uncertainties reflecting, in ag- gregate, the changing costs and benefits of all generations and end-users. There have been worldwide moves over the last two decades to restructure electricity industries so that generators and end-users see price signals that more appropriately reflect these underlying industry economics. In their simplest form, FIT schemes can effectively shield project developers from such energy market signals through a fixed payment for each MWh of renewable generation independent of the value it ac- tually provides for the industry at that time and location within the network [2]. Simplified tendering processes awarded to projects on the basis of lowest required government payments per MWh of renewable generation, which were adopted in countries such as Ireland and China, can have similar impacts. Other policy approaches such as renewable electricity pro-

duction taxcreditsas seen in theU.S., and tradablegreencertifi- cates as seen in a number of European countries and Australia, provide another approach for supporting wind energy. By com- parison, these can ensure that wind farm developers and opera- tors are still incentivizedbyelectricitymarket “signals” tomax- imize overall industry value. The last few years has seen important developments in a

number of countries that can help us better understand these issues. For example, Denmark and Spain have moved from a conventional FIT to a tariff premium above the electricity market price, the latter with additional arrangements that cap potential incomes to wind generators. The UK Renewables Obligation scheme now appears to be driving greater industry development, especially in offshore projects. The U.S. Federal production tax credits and state-based renewableportfolio stan- dardshavealsodrivenvery significant if sometimesboom–bust winduptake,particularly inTexaswithaquarterof thatnation’s installed capacity. Table I shows the total wind energy installed capacity at the end of 2010 in the regions considered in this paper together with their proportion of electricity consumption now supplied by wind energy. Wind generation penetrations have now reached significant

levels (from 10%–20%) in countries such as Denmark and Spain, and states such as South Australia and Iowa. This, in turn, has driven changes in electricity market design and wider policyarrangements in thesecountries inorder tobettermanage the major contributions of highly variable and only somewhat predictablewindgenerationwithin theirpowersystems.Formal participation by wind generation in electricity market dispatch and ancillary services may be limited to day-ahead markets

1949-3029/$31.00 © 2012 IEEE

810 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

TABLE I TOTAL INSTALLED WIND ENERGY CAPACITY AT THE END OF 2010 AND THE PROPORTION OF ELECTRICITY CONSUMPTION SUPPLIED IN 2010

FOR SELECTED COUNTRIES AND REGIONS

or include real-time markets and even ancillary services such as voltage and frequency control. Improved wind forecasting systems have reduced prediction errors, whereas delayed gate closures and active demand participation have decreased the potential energy imbalances that have to be resolved by the industry. Other electricity market arrangements that may affect wind energy are charges due to imbalances settlement, addi- tional income due to capacity recognition and rewards when wind generators reduce their output following orders from transmission system and market operators. This paper draws together someof thekeyexperiences, chal-

lenges, and responses to growing wind penetrations in selected jurisdictions within Europe, the United States, and Australia. These are by no means the only countries from which lessons might be drawn, or where significant wind industry develop- ment is underway. However, they do represent important and interesting examples of some key current and possible future wind markets, and the range of support policy approaches and electricitymarket arrangements thatmaybeemployed to facili- tate highwindpenetrations.The selected states inAustralia and theU.S.are thosewith thehighest installedwindcapacities.The paper is divided into four further sections. Section II presents the main support policies in place for wind in these selected ju- risdictions. Section III provides an overview of their different electricity market arrangements. Section IV discusses the inter- actions between wind energy and the support policies and elec- tricity markets considered in the previous two sections. Finally, conclusions are presented in Section V.

II. SUPPORT POLICIES

A wide range of policy mechanisms to support wind energy, or renewable energy more generally, have been used by dif- ferent jurisdictions over recent decades. A general assessment of the available support policy mechanisms and their potential strengthsandweaknessescanbefoundin[3].Fourgeneralwind energy support policy mechanisms are considered here. The ju- risdictionscovered in thispaper thathaveoptedforeachof these four approaches, together with their particular characteristics, are described below and summarized in Table II.

TABLE II RENEWABLE ENERGY SUPPORT POLICIES IN DIFFERENT REGIONS

A. Tender Schemes

The theoretical basis of tender schemes is highly promising—governments can set a target of installed capacity or total public expenditure, and invite prospective project de- velopers to submit project tenders that specify the government support—capital $ or $/MWh—required to proceed. Gov- ernments can then choose the lowest cost project providers. Unfortunately, the experience to date with tender-based ap- proaches is mixed. For example, a number of countries opted for this approach in order to drive initial deployment of re- newable energies but abandoned it some years later. In 1990, the UK introduced the Non Fossil Fuel Obligation mainly as a policy to support the nuclear industry although it also drove the installation of a number of wind farms. Ireland introduced its tender scheme in 1996 based on the UK model, but abandoned it ten years later because it failed to reach its set targets [4]. It is also notable that China adopted a franchise tender program in 2003 that was abandoned in 2009 for onshore projects. What proved to be irrationally low bids offered by competing developers led to the Government selecting projects with such lowrequired support prices that the“winning”proponentswere later unwilling or unable to actually undertake their projects [5]. However, China kept a tender scheme for offshore wind

APARICIO et al.: COMPARISON OF WIND ENERGY SUPPORT POLICY AND ELECTRICITY MARKET DESIGN 811

farms and Denmark opened a scheme in2005, also for offshore wind, with very good results [6]. A number of U.S. states use tender-based processes for their

renewable portfolio standard although these may also be based around the use of tradable certificates as outlined in the next section.The jurisdictionsusing such tendershavealsoachieved mixed success [7].

B. Feed-in Tariffs

Feed-in tariff schemes adopted by Denmark, Germany, and Spain have without doubt been the primary drivers of their significant wind energy deployment over the last two decades. Initial policy settings in these three countries, announced in the 1990s,were similar and relatively simple schemeswith a single tariff for all renewable producers. These FIT schemes provided an electricity consumer funded fixed price for each MWh of generation over a given time period. Any project meeting the scheme requirements was eligible for this payment. As wind installed capacity started to rise, however, these policies have been significantly amended. Each country has followed different strategies. Germany has decided to make important changes in the FIT

scheme, includingfixeddegression,anequalizationschemethat tries to compensate the differences in wind resources between regions, and higher tariffs for repowering and offshore wind farms.The latest amendment of theGermanRenewableEnergy Act, in forcesinceJanuary1,2012,has increased thedegression for both onshore and offshore projects. However, the reduction inoffshore tariffswillnotbeapplicableuntil 2018 insteadof the originally proposed 2015. An optional accelerated repayment model which offers a higher initial tariff for a reduced number of years has also been introduced for offshore wind farms. This amendmentalso introducedanewmarketpremium,openingthe possibility fordirect selling(ordirectmarketing),whichhas im- portant implications for the generators that participate, as it is shown latter in the paper. German amendments have managed tokeepannual increases in installedcapacity relativelyconstant over the past decade as Fig. 1 shows. Denmark phased out its FIT scheme in 2000. After a transi-

tionperiod, it adoptedaschemewhere the“feed in” tariff isnow afixedpremiumpaymentaboveandbeyondwhat thewindfarm projects earn from the electricity market. In Denmark’s case, wind generators connected to the grid after January 2003 must sell their production to the electricity market. Fig. 1 shows that new installed capacity rapidly declined following this change. However, andasnotedabove, the tenderingprocess foroffshore wind farms has driven more than 200 MW of new capacity in both 2009 and in 2010, and in 2013 the Denmark’s largest off- shore wind farm with 400 MW will start operation and is ex- pected to supply 4% of the country’s demand. This will help to meet the Danish target of 50% from wind by 2020. Spain introduced theoptionofwind farms takinganFITpre-

mium in addition to energy market prices in 1998, earlier than Germany, but also kept the conventional FIT mechanism. Thus wind energy producers are able to choose between both remu- neration schemes and switch between them every 12 months. No wind energy producers initially decided to participate in the

Fig. 1. Annual increases in installed wind capacity in Spain, Germany, and Denmark from 2002 to 2011. Sources: Respective national wind associations.

electricity market so an amendment in 2004 provided extra in- centives to switch. By the end of 2006, around 90% of wind generationcapacitybid in themarketas thosearrangementspro- vided higher revenues than the conventional FIT. Finally, fur- ther amendments, announced in 2007, modified the tariffs in- troducing a cap and a floor in the sum of market price plus pre- mium, variable degression depending on inflation, and lower tariffs once a technology target has been reached. Many wind energy developers accelerated the installation of their projects in order to complete them before this amendment came into ef- fect at the beginning of 2008. The year 2007, therefore, saw a record increase in installed capacity, as shown inFig. 1.Thede- crease in new installed capacity over the last two years is due to the“pre-assignation” register introducedby theSpanishgov- ernment in 2010. A limited number of projects are approved in order to ensure Spain does not surpass its targets for the wind. Moreover, in January 2012, the government announced a tem- porary moratorium that freezes policy support for any new re- newable energy project due to the impacts of the Global Finan- cial Crisis. Ireland and China have now replaced their tender schemes

withFITs.InIreland,FITshavebecomethemainmechanismfor supporting wind energy. Offshore wind farms have had higher tariffs since2008 [4]. InAugust 2009,China announced itsfirst FIT scheme for onshore wind with different tariffs that depend on thewind resources and investment conditions in eachof four regions [8].

812 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

C. Quota System/Tradable Certificates

Aquota systemisusually relatedwith tradable renewableen- ergy certificates (or credits) and has different names depending on the country. In the UK, the scheme that came into force in 2002 is known as Renewables Obligation (RO). Wind genera- tors receivedaRenewableObligationcertificate (ROC)foreach MWh of electricity generated. A 2008 amendment introduced the concept of ROC banding in order to give additional ROCs to emerging technologies. Thus, onshore wind, included in the “reference” band, receives 1.0 ROC per MWh, while offshore wind, which is included in the “postdemonstration” band, ini- tially received 1.5 ROCs per MWh. The 2009 budget raised it to2ROCsfor2009/10and1.75 for2010/11.TheROCsaresold to electricity suppliers in order to fulfil a mandated obligation placed upon them by the government according to its national renewable energy quota or target. Suppliers can either present enough ROCs to cover their obligations or they can pay for any shortfall into a buyout fund. The Renewables Obligation Order 2009 introduced significant changes. It requests the Secretary of State to announce the obligation level six months preceding an obligation period. This obligation level is the greater value of either the number of ROCs needed to meet a fixed target of ROCs/MWh or a headroom that is calculated as the ROCs ex- pected to be issued according to the amount of renewable elec- tricity expected to be generated, uplifted by 10%. This “guar- anteed headroom” mechanism sets an effective floor for ROC prices once the obligation is reached. In the case where ROC supply exceeds theobligateddemand, therefore, theROsystem will then effectively operate in a similar manner to an FIT pre- mium mechanism. TheAustralianGovernment’sMandatoryRenewableEnergy

Target (MRET) commenced operation in 2001 as the world’s first renewable energy certificate trading scheme [2]. It requires all Australian electricity retailers and wholesale electricity cus- tomers to source an increasing amount of their electricity from newrenewablegeneration sources.The liablepartieswithin the Australian electricity market are electricity retailers and those large consumers who purchase directly from the wholesale market. The “additional renewable electricity” that the liable parties are required to acquire was originally intended to be equivalent to 2% of their electricity purchases by 2010. The Renewable Energy Target (RET) announced in 2009 raised the requirement to 20% by 2020. Targets to date have been easily met and the costs seem reasonable by international standards [2]. Some State Governments have also set jurisdictional renew-

able goals although they are not backed with specific obliga- tions. For example, in 2007, South Australia set a 20% target for2014. In June2011, it hasbeenalreadymet.Thestate, there- fore, set a new goal of 33% by 2020 [9]. Victoria, which has the second highest installed wind capacity, has a 2020 goal of 25%, with a minimum of 20% for wind energy. However, it is intended that the Federal RET provide almost all wind project support in Australia. A number of changes have been made to the schemeover its decadeof operationbeyondagreater target, including separate arrangements for large-scale and small-scale renewable energy systems [2].

Currently, there isnoafederalquotaschemein theU.S.How- ever, 29 states, the District of Columbia, and Puerto Rico have a Renewable Portfolio Standard (RPS), and eight states have a renewable portfolio goal. Texas adopted both an RPS and a renewable energy credit

(REC) trading program in 1999 [10]. The RPS target was 2000 MW of new renewable generation by 2009, in addition to the 880 MW installed at the time. It was raised in 2005 to 5880 MW by 2015, where 500 MW must be resources other than wind, and to 10000 MW by 2025. According to the 2009 compliance report, Texas had already surpassed its 2025 target by 2009. The Electric Reliability Council of Texas (ERCOT) acts as the program administrator of the REC trading program. Iowa passed one of the earliest renewable energy laws in

the U.S. in 1983. It allocated 105 MW of renewable gener- ating capacity between the two Iowan investor-owned utilities: Mid-AmericanEnergyCompany (MEC)andAlliantEnergy In- terstate Power and Light (IPL) [10]. As Table I shows, the re- quirementhasbeenclearly surpassed sobothutilitieshavebeen authorized to export RECs by participating in Midwest Renew- ableEnergyTrackingSystem, Inc. (M-RETS).Renewablegen- eratorsusedformeetingtheRPSarenotallowedtoexportRECs in order to avoid double-counting [11]. A voluntary goal of 1000MWofwindenergycapacityby2010,establishedin2001, has also been easily exceeded. Section 476.53 in Code of Iowa (2009) provides that it is the intent of the general assembly to attract the development of electric power generating facilities within the state. Thus, when eligible new electric generation is constructed by a rate-regulated public utility, the Iowa Utilities Board, upon request, must specify in advance the ratemaking principles that will apply when the costs of the new installa- tionare included in electricity rates. InMarch2009, pursuant to section 476.53, MEC filed an application for determination of advance ratemakingprinciples for up to1001MWofnewwind generation to be built in Iowa from 2009 through 2012. In De- cember 2009, the Board took up MEC’s proposal. As a result, MEC has installed significant wind generation capacity and is expected to have a total 2284 MW by the end of 2012. Minnesota introduced twoseparateRPSpolicies in2007,one

for theutilityXcelEnergyand the second for other electric util- ities. The latter includes public utilities providing electric ser- vice, generation, and transmission cooperative electric associa- tions, municipal power agencies, and power districts operating in the state [10]. The RPS for Xcel Energy requires 30% of its total retail electricity sales in Minnesota to come from renew- able sources by 2020. It included a minimum of 25% for wind energybut aStateSenateBill passed in2009addedamaximum of 1% from solar to this requirement. Thus, at least 24% must come from windenergy, up to1% maycome from solar energy, and the other 5% may come from other eligible technologies. TheRPSforotherutilities requires25%of their total retail elec- tricity sales in Minnesota to come from renewable sources by 2020 without any technology minimums. Minnesota has been included in the M-RETS since 2008. This tracking system pro- gram ascribes the same amount of credits to all eligible tech- nologies independently of the state where electricity is gener- ated.XcelEnergyisnotallowedtosellRECstootherMinnesota utilities for RPS-compliance purposes until 2021.

APARICIO et al.: COMPARISON OF WIND ENERGY SUPPORT POLICY AND ELECTRICITY MARKET DESIGN 813

California launched an RPS in 2002 with a target for its electric utilities to have 20% of their retail sales derived from eligible renewable energy resources in 2010 [10]. Senate Bill X1-2, enacted in 2011, raised the requirement to 33% by 2020. The Bill also established three categories, also known as buckets, of RPS-eligible electricity applicable to contracts executed from June 2010. The decision adopted in December 2011 by the California Public Utilities Commission provides detailed requirements for the three categories. Category one is for electricity that is from an RPS-eligible generation installa- tion that has its first point of interconnection with a California Balancing Authority (CBA); scheduled from an RPS-eligible generation installation into a CBA without substituting elec- tricity; or dynamically transferred to a CBA. Category two is for electricity that is firmed and shaped, providing incremental electricity scheduled into a CBA. Category three is for those transactions that do not meet the criteria of any of the two previous categories, including unbundled RECs. There are limitations on the amount of generation procured in categories two and three that become increasingly narrow over time. Categoryonegeneration is required tobeaminimumof50%of the total for the compliance period ending in 2013, 65% for the compliance period ending in 2016, and 75% thereafter. Until the portfolio content categories become sufficiently clear, the utilities are preferencing power purchase agreements (PPAs) with installations that definitely belong to category one. Given California’s increasing target, it is envisaged that PPA opportu- nities in categories two and three will expand in time, and with greater clarity on the arrangements. The Western Renewable Energy Generation Information

System (WREGIS) tracks the renewable energy generated in the region covered by the Western Electricity Coordinating Council (WECC), California included. WREGIS issues cer- tificates for every REC generated, which can be used to verify compliance with state RPS. However, currently WREGIS is not able to track the three new portfolio content categories.

D. Tax Credits

The U.S. Federal production tax credit (PTC) for wind has had a checkered history over the past decade. The latest exten- sionof2009providesan incometaxcreditof2.2¢/kWhuntil the endof2012while addinganumberofprovisions.Taxpayers el- igible for the PTC are allowed to take a business energy invest- ment tax credit (ITC) equal to30% of the construction costs for the installation or to receive a cash grant of equivalent value if construction began by the end of 2011. Before 2009, installa- tion owners that did not generate enough taxable income were unable to utilize PTC so they had to monetize the PTC through tax equity investors. The PTC, which applies for the first ten years of electricity

production [12], has been remarkably successful in supporting wind deployment when it has been in place. However, during the periodic lapses of the PTC prior to congressional renewal, the state-based RPS mechanisms alone were not able to sustain the growth of wind power [13]. As it represents a credit against passive income, the PTC has a significant resemblance to an FIT premium. In fact, both are a fixed cash incentive provided to each kWh generated by wind.

Finland is the only EU country which uses tax incentives as the main support scheme for renewable energies. This policy, however, has not been effective for wind development [6].

III. OVERVIEW OF ELECTRICITY MARKETS AROUND THE WORLD

Jurisdictionsaround theworldhave takenawide rangeofap- proaches to electricity industry restructuring over the past three decades. Care must, therefore, be taken when comparing in- dustry approaches and performance. Generally, countries with restructured electricity industries have both forward markets and real-time markets. In the forward markets, electricity is traded either centrally on a power exchange or bilaterally di- rectly between market participants. A key role for the forward markets is to support unit commitment of those thermal genera- tors that require generation scheduling aday prior to energyde- livery. In theday-aheadmarkets,electricity is traded in intervals (settlementperiods) thatmaybeonehour longor lessdepending on the market design. A key challenge for these arrangements is that unexpected generator outages or changes in demand can take place between the closing of the day-ahead market and de- livery thenextday. Intradaymarketspermitmarketparticipants to trade closer to thedelivery time, upuntil just prior to thegate closure. Intraday markets are commonly a continuous trading market that operates with a gate closure set one hour ahead of the settlement period [14]. After gate closure, it is no longer possible to change bids and offers for the settlement period. Fromgateclosureuntil real-time, thedifferencebetweensupply and demand is continuously balanced through real-time mar- kets,where transmissionsystemoperators (TSOs)—orindepen- dent system operators (ISOs) or regional transmission organi- zations (RTOs)—purchase the energy needed to match supply to demand and to solve any network constraints. These orga- nizations also have a responsibility for imbalance settlement. Chargesapply togenerators if theirenergydeliveriesdiffer from theoffers submitted to themarket.Thepayment dependson the price system. This may be a two-price system, which has one price for imbalanceswith thesamesignas thesystemnet imbal- ance (prejudicial for the system since they contribute to the net imbalance)andadifferentprice (lower) for imbalanceswithop- posite sign (beneficial since they counteract the net imbalance); or a one-price system, which has only one price for all imbal- ances. Normally, the two-price system is used, since it encour- ages market participants to limit their imbalances according to whether they add to, or subtract from, net system imbalances. Toensurepower systemsecurityand reliability, ancillaryser-

vices are needed, including frequency and voltage control and black-start capability. TSOs or their equivalent organizations purchase ancillary services from service providers. Frequency control ancillary services are commonly tradedonamarket that has marked similarities to the electricity market. Gate closure, the duration of settlement periods, and imbal-

ance settlement arrangements are all potentially very important for wind energy integration. The brief descriptions of elec- tricity market arrangements for each of the jurisdictions that we are considering here, therefore, pay special attention to these features.

814 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

A. Denmark

The four countries comprising the Nordic region (i.e., Den- mark,Sweden,Norway,andFinland)areamongthefirst tohave restructured their electricity industries. In 1993, they connected their individual markets creating Nord Pool Spot, which was the world’s firstmultinational power exchange. Nord PoolSpot trades 74% of the electricity generated in the Nordic region. The rest is traded through bilateral contracts. It also operates a day-ahead market called Elspot, and an intraday market called Elbas. Elspot trades in one-hour intervals and closes at 12:00.1

It introduced different area prices in order to deal with network congestion between countries and within Norway [15]. Elbas is a continuous intraday market that covers the Nordic Region, Germany, and Estonia with a gate closure one hour before de- livery. This intraday market is becoming increasingly signifi- cant as more wind energy enters the grid given that imbalances between its day-ahead contracts and produced volumes often needtobeoffset [15].Asdiscussedbelow, thesemarketarrange- ments are now being changed in ways that affect wind energy directly. Since April 2011, the gate closure for trading in Germany

has been reduced to 30 min. Negative prices have also been introduced inallmarket areas sincemid-February2011 inorder to price oversupply [16]. Each local TSO has a different set of ancillary services costs

according to their procurement arrangements and reserve re- quirements. Energinet.dk is the Danish TSO and purchases dif- ferent ancillary services inWesternandEasternDenmarkas the former is synchronouslyconnected to theUCTEsystemand the latter to the Nordel system. Denmark settles imbalances using a two-price system. Norway used one-price system until 2008. Since then, all Nordic countries have used a two-price system forproduction imbalances andaone-price systemforconsump- tion imbalances [17].

B. Spain

Spainbelongs,withPortugal, toMercado Ibéricode laElect- ricidad (MIBEL)—the IberianElectricityMarket.Eachcountry may have different prices (splitting) in the case of transmis- sion restrictions. All market participants can either arrange bi- lateral physical contracts or participate in a day-ahead market. As with the Danish market, this day-ahead market is divided into one-hour settlement periods and closes at 12:00. After the day-aheadscheduling,generatorpositionscanbeadjusted in the intradaymarket.Rather thanbeingacontinuousmarket, it is di- vided into six sessions.Eachsessionhasadifferentgate closure (around2hours)aswell asadifferent timeofdeliveryscope(up to 9 hours in the sixth session). The provision of ancillary services follows the general rules

common in theUCTEsystem.Theprimarycontrol is a compul- sory service shared across all generators and without remuner- ation. The secondary control is performed within control areas according to the requirements, in MW, set by the Spanish TSO (Red Eléctrica de España). The generators willing to offer this

1This time of day and the following ones are in 24-h notation.

service bid their power available and a market clearing process calculates the price per MW. The price is paid even if their ser- vicesarenot required(availability).The tertiarycontrol restores the secondary control reserves under emergencyconditions and while it is optional for generators to formally participate, the TSO can call upon any generator should it be required. By con- trast with secondary control reserves, tertiary control services areonlypaid for ifused. Imbalancesare settledwitha two-price system.

C. Germany

Most wind generation is not scheduled and, instead, is intro- duced into the electricity market through one of the country’s fourTSOs.Thedistributionnetworkoperators transfer thewind generationforafixedprice to their respectiveTSO,which trans- forms the load fluctuating profiles into standard load profiles which are sold to all utilities [18]. Customers pay an average tariff to utilities. According to [18], this profile transformation mechanism isnot fully transparent andnot cost-optimized.Fur- thermore, as the costs of the profile transformation are com- pletelypassedthroughtothenetworkcustomers, there isnoeco- nomic incentive to minimize these costs. This mechanism has thereforebeenargued to represent an importantweaknessof the GermanFITscheme.However, theamendmentapplicable from January2012 [19]mayhelp inaddressing thesedrawbacks.Re- newablegenerators candecideonamonthlybasis to change the remunerationmechanismsandparticipate indirect selling in the electricity market. They have to forecast their production and are directly charged for their imbalances. The additional costs are covered by an extra premium known as management pre- mium. TSOssharewindenergy imbalances,which reduces theneed

for reserves. In order to manage this equitably, it is allocated proportionally to theTSOs’according to their consumption,not their installed wind power.

D. United Kingdom

The British Electricity Trading and Transmission Arrange- ments (BETTA) cover England, Wales, and Scotland. Northern Ireland has been part of the Single Electricity Market together with the Republic of Ireland (see Section III-E) since 2007. In BETTA, over 90% of electricity is traded through unrestricted bilateral contracts. A power exchange permits market partici- pants tofine tune their contractedpositions.Gate closure is cur- rently set one hour ahead of each half hourly settlement period. ELEXON, the Balancing and Settlement Code Company, uses a two-price system [20].

E. Ireland

The Single Electricity Market (SEM) commenced trading in IrelandandNorthern Irelandonanall-islandbasis inNovember 2007. SEM is a mandatory power exchange where all Ireland’s electricity must be traded. It has only a day-ahead market with a gate closure at 10:00. Energy is settled weekly. SEM plans to develop an intraday market [21].

APARICIO et al.: COMPARISON OF WIND ENERGY SUPPORT POLICY AND ELECTRICITY MARKET DESIGN 815

F. Australia

The Australian National Electricity Market (NEM) includes all states and territories other than Western Australia and the NorthernTerritory. Itscenterpiece isasetof regionalgross-pool spot energy and ancillary services markets that solve a secu- rity-constrained dispatch every 5 min. The Australian Energy MarketOperator (AEMO) is thewholesalemarket operator and TSO for the entire system. Regions are currently located at all borders between states within the NEM. All generating plants of greater than30-MWcapacity (except intermittent generation including wind) are required to participate as scheduled gen- erators and submit offers to sell or bids to buy energy (and/or ancillary services) in the NEM dispatch process. The predis- patch processes forecasts up to 40 hours ahead of real time and provides public forecasts of energy and ancillary service prices and (privately to each dispatchable participant) dispatch levels based on participant bids and offers, the demand forecasts and the estimated effects of dispatch constraints. Demand is per- mitted to participate directly in the wholesale market; however, nearly all end-users interface with the market through an elec- tricity retailer [2]. There are eight Frequency Control Ancillary Services

(FCAS) markets to provide load following (raise and lower) and three contingency responses of different speed (raise and lower) between the 5-min energy dispatches. Market dispatch co-optimizes energy and FCAS bids and offers to establish re- gional prices for both energy and FCAS for each 5-min period. Commercial trading is based on these prices averaged over 30 min. Locational pricing within regions is achieved using averaged loss factors. Importantly all generators are permitted to change their offers (rebid) just prior to each 5-min dispatch. Furthermore, the only commercially significant prices in the NEM are these averaged30-minprices—the predispatchprices are advisory only. Note also that the NEM is an energy-only market and participants are required to manage their own unit commitment and other intertemporal scheduling challenges (within a range of technical dispatch constraints).

G. Texas

ERCOT manages the electricity industry arrangements sup- plying 85% of Texas demand and covering 75% of state land area. The ERCOT control area is not synchronously connected to either the Eastern or Western Interconnection. However, it can exchange about 860 MW through dc links. In December 2010, ERCOT switched from a zonal market to a nodal market in order to improve price signals and dispatch efficiencies and assign localcongestiondirectly[22].Theday-aheadmarketem- ploys a co-optimization engine that uses both energy and ancil- lary services offers to calculate the energy schedules and ca- pacity awards. ERCOT closes this market at 14:30. The real-timemarket is called security constrainedeconomic

dispatch(SCED).ERCOTgenerally runs theSCEDevery5min using offers by individual resources and actual shift factors by each resourceoneach transmissionelement.The settlementpe- riods are 15 min long.

H. Minnesota and Iowa

Minnesota and Iowa belong to the Midwest ISO (MISO), which operates a day-ahead market and a real-time and op- erating reserves market. They coexist with both financial and physical bilateral transactions between industry participants. The day-ahead market simultaneously clears energy and op- erating reserves on a co-optimized basis for every one-hour settlement period. Security constrained unit commitment (SCUC) and SCED algorithms ensure the scheduling of ade- quate resources [23]. The real-time and operating reserves market uses an SCED

algorithm to simultaneously balance supply and demand and to meet operating reserves requirements amongst other actions. The gate closure is set to only 30 min ahead of delivery. InMarch2011,an importantchangecame intoeffectwith the

creation of a new category of resources called Dispatchable In- termittentResources (DIRs).Thiscategoryonlyapplies towind farmsandallows themtovoluntarilyparticipate in the real-time market fromJune2011,where theyareeligible tosupplyenergy but not operating reserves. From September 2011, DIR imbal- ances are settled similarly to conventional generators although only when an 8% tolerance band is exceeded, with a minimum of 6 MWh and a maximum of 30 MWh, for four or more con- secutive 5-min intervals within an hour. Wind generators are exempt of imbalance charges in cases of force majeure, such as extreme winds.

I. California

The California ISO (CAISO) has three day-ahead processes: a market power mitigation determination, integrated forward market, and residual unit commitment. A bid from a market participant that fails the market power test is automatically re- duced to the reference level price of that participant, and the system determines the minimal and most efficient schedule of generation to address local reliability. The integrated forward market simultaneously analyzes the energy and ancillary ser- vices market to determine the transmission capacity required (congestion management) and confirm the reserves that will be needed to balance supply and demand based on supply and de- mand bids. It ensures generation meets load and that all final schedules are feasible with respect to transmission constraints as well as ancillary services requirements. When forecast load is not met in the integrated forward market, the residual unit commitment process enables CAISO to procure additional ca- pacity by identifying the least cost resources available [24]. The real-time market produces energy to balance instanta-

neous demand, reduce supply if demand falls, offer ancillary services as needed and, in extreme conditions, curtail demand. The gate closure is set 75 min ahead of delivery. The market has two unit commitment mechanisms. Real-time unit commit- ment assigns fast- and short-start units in 15-min intervals and looks forward 15 min, while short-term unit commitment as- signs short- and medium-start units every hour and looks for- ward three hours beyond the settlement period every15min. In real-time, the economic dispatch process dispatches imbalance volumes and energy from ancillary services. It runs automat- ically and dispatches every 5 min for a single 5-min interval.

816 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

Under certain contingency situations, CAISO may dispatch for a single 10-min interval.

IV. WIND ENERGY PARTICIPATION IN ELECTRICITY MARKETS

A compilation of some key themes for wind energy integra- tion coming out of the individual jurisdictional experiences is presented as follows.

A. Wind Energy Forecasts and Generation Scheduling

Windgeneratorsparticipatinginday-aheadmarketsmustpre- dict theiroutputbeforemarketclosetime(varyingbetween9.5h in ERCOT to 40 h ahead of delivery in some other markets). Theday-aheadand longerwind forecasts required for suchgen- eration scheduling are not sufficiently precise and two possible alternatives have been put forward to help manage this. In the first case, adopted in Germany (not for renewable generators that participate in direct selling in the electricity market), Aus- tralia, MISO, ERCOT, and CAISO, the output of all wind gen- erators is predictedover a rangeof timehorizons throughacen- tralized forecasting system. In the second approach, adopted in Denmark, Spain, and the UK, numerous prediction companies compete to provide forecasts for wind farm clients [25]. Note that such wind energy forecasts have value to all market partic- ipants, not just the wind farms—another argument for centrally provided forecasting services. Even in industries where wind is not required to partici-

pate in forward markets there is considerable value in wind farms and other market participants having useful day-ahead forecasts. These can play a role in derivative market trading around the future spot price, maintenance scheduling of wind farms, unit commitment strategies of thermal plant, and pro- duction scheduling of hydro. There is also considerable value in useful short-term forecasts that assist in generator bidding in the real-time spot markets. For example, recent changes in the Australian market arrangements have more formally incorporated wind farms into market scheduling and ancillary services arrangements through a semischeduled classification. There are similarities with the new DIR category within the MISO arrangements that was described in the previous section.

B. Imbalance Settlement

Imbalance settlement is probably the aspect of electricity markets design that has highest impact on wind energy [17], [26]. Ingeneral, systemandmarketoperatorscalculate such im- balances and settle them according to either a one- or two-price system. This determines how the total balancing costs are distributed and how incentives are given to market participants [27]. With the one-price system used for balance settlement in Norway before 2008, the balance costs of wind energy were negligible as long as its random variability assured that positive imbalances from some wind generation were compen- sated by negative ones [28]. Since 2008, all Nordic countries have used a two-price system for production imbalances and a one-price system for consumption imbalances [17]. Spain originally applied a different one-price system that charged for

all imbalance volumes independently of their direction. Now it uses this two-price system: zero costs for producers that do not contribute to the system net imbalance and a penalty for those that do. Denmark, the UK, and Australia (only with Regulating FCAS)also runa“causerpaysprocedure” inorder to assign the cost of the regulating power to those market participants who are responsible for the imbalance. Note, however, that there are inevitably difficulties in assigning such responsibilities given the complex nature of electricity industry operation. This kind of arrangement may impose significant charges on wind energy so some tolerance in energy imbalance is often applied and the prices paid by wind are often lower than those paid by conventional generation. In Germany, full responsibility for balancing wind genera-

tion is assumedby theTSOs. Indeed, they are in chargeof fore- casting, scheduling, and balancing. In the U.S., Federal order 890assists intermittent generationwithmoreflexiblebalancing settlement [29]. For example, DIRs in MISO are only charged if an 8% tolerance band is exceeded for four or more consec- utive 5-min intervals within an hour, whereas CAISO has the Participating IntermittentResourceProgram(PIRP).Windgen- erators that participate in the PIRP have better arrangements. Energy imbalances are netted on a monthly basis and settled at a monthly weighted market-clearing price. InAustralia, thehybrid5/30mingross spotmarketwithasso-

ciated frequency control ancillary services (FCAS) seems rea- sonably supportive of wind integration [2]. Reduced gate closures permit rebidding up to few minutes

from delivery whereas in many intraday markets it is possible to reschedule with updated forecasts of only one hour (or less as in MISO). However, this can reduce liquidity; especially in the case of market power (CAISO has market power mitiga- tion). If themarket is openuntil very close togate closure,wind generators are able to better forecast and manage the energy that they actually produce. However, the closer to real time, the fewer the conventional generators that may be available to help wind generators to correct their position. Continuous mar- kets permit changes to bids closer to real time (for example, in the UK until one hour before). In Spain, by contrast, the six in- traday sessions have delivery scopes up to 9 hours. There are higher forecast errors; however, there is also a greater willing- nessand interestamongst thegenerators tochange theirposition (including all wind generators), so the market is likely to have greater liquidity [18]. Experience to date suggests that intraday pricesdiffer little fromtheday-aheadprices.So thismore liquid market permits generators to improve their bids and offers at littlecostwhilealsoreducingregulationrequirements.Note that gross pool arrangements such as those of the Australian NEM resolveshort-termsupply–demandbalancewith thecompulsory involvement of all generation and load. Wind imbalances are significantly reduced by aggregating

the bids of wind farms over geographically dispersed locations and large areas [26]. Active demand participation is also useful in reducing imbalances. FERC Order 719 considers electricity market accepting bids from demand response resources, on a basis comparable to any others, for ancillary services that are acquired in a competitive bidding process.

APARICIO et al.: COMPARISON OF WIND ENERGY SUPPORT POLICY AND ELECTRICITY MARKET DESIGN 817

C. Curtailment

An excess of wind generation may cause system operating problems such as transmission line overloading or insufficient regulation reserves that force TSOs or their equivalent organi- zations toorder real-timecurtailment towindgenerators during normal operation. Wind generators that reduce power may be rewarded for this, depending on the electricity market arrange- ments. When rewards are given, wind generators typically earn a percentage of what they could have generated. In Ireland it is 100% [30] while in Spain it is just 15%. Another possibility to reduce over-production is to permit negative wholesale market prices [27]. Prices in electricity markets typically have a zero-floor limit. Elbas has accepted negative prices since 2009 while Australia and some U.S. electricity markets have permittednegativepriceswellbefore this.CAISOhasproposed to lower the bid floor from $30/MWh to $150/MWh, then to $300/MWh. Spain may consider accepting it only for downward regulation provision.

V. CONCLUSION

Wind generation penetrations have reached significant levels in some countries and states around the world. In almost all cases this has required changes in both policy support mech- anisms and electricity market design in order to better manage wind energy. Tender schemes have been successful only with offshore wind. FIT tariff premiums are now the predominant scheme as a market oriented transition from conventional FITs. Denmark made the transition mandatory whereas Spain, and Germany since 2012, introduced incentives. For the case of Spain, theypersuadedaround90%ofwindgenerators to switch from FITs. The “guaranteed headroom” mechanism introduced in the UK is a way to transform the Renewables Obligation mechanism into FIT premiums given an oversupply of ROCs. The PTC can also be considered as a form of FIT premiums. Quota-based approacheshavehadmore limited applicationand mixed success to date. With regard to electricity market arrangements, imbalance

settlement is probably the element of electricity market design thathas thehighest impactonwindgeneration. Itdependsonthe price system, the specificarrangements for intermittent sources, possibilities to aggregate bids of wind farms over geographi- cally dispersed sites and large areas, active demand participa- tion, and gate closure. Small gate closures permit scheduling with reduced forecast errors. However, this reduces liquidity. In conclusion, there are complex and changing interactions

between a) the desired policy objectives of increasing wind energy generation or renewable energy more generally, b) the chosen policy approaches applied to facilitate greater deploy- ment, and c) the commercial and regulatory arrangements that govern how such wind energy is integrated into existing elec- tricity industries. It is evident that the challenges of appropriate policy and electricity market arrangements grow as wind pen- etrations increase. Some clear trends have emerged with those countries now experiencing high penetrations. These include the need to have wind farms more formally participating in the electricity market mechanisms that manage supply-demand balanceover the immediate to longer-term.While thismight be

seen as an impediment to greater wind deployment, it is better understood as the inevitable process of wind transitioning from a small industry contributor that can be ignored as “negative load,” to a seriousplayer that cangreatly help in addressingour growing energy security and climate change challenges within the electricity sector.

REFERENCES [1] REN21, Renewables 2011 Global Status Report 2011. [2] I.MacGill, “Electricitymarket design for facilitating the integrationof

wind energy: Experience and prospects with the Australian National Electricity Market,” Energy Policy, vol. 38, pp. 3180–3191, 2010.

[3] N. Enzensberger, M. Wietschel, and O. Rentz, “Policy instruments fostering wind energy projects—A multi-perspective evaluation approach,” Energy Policy, vol. 30, pp. 793–801, 2002.

[4] European Renewable Energy Council, Renewable Energy Policy Re- view Ireland, Mar. 2009.

[5] F. Wen, D. Hua, Q. Wang, and S. N. Singh, “Wind power generation in China: Present status and future prospects,” Int. J. Energy Technol. Policy, vol. 6, pp. 254–276, 2008.

[6] The Support of Electricity From Renewable Energy Sources Commi- sion staff working document; SEC(2008) 57, Jan. 2007.

[7] K. Cory, T. Couture, and C. Kreycik, Feed-in Tariff Policy: Design, Implementation, and RPS Policy Interactions NREL, Tech. Rep. NREL/TP-6A2-45549, Mar. 2009.

[8] Z.-Y. Zhao, J. Zuo, L.-L. Fan, and G. Zillante, “Impacts of renewable energy regulations on the structure of power generation in China—A critical analysis,” Renew. Energy, vol. 36, pp. 24–30, 2011.

[9] South Australia Meets 20% Renewable Energy Target, Ministerial Statement Jun. 2011.

[10] DSIRE, Database of State Incentives for Renewables & Efficiency [Online]. Available: http://www.dsireusa.org

[11] Order Approving Facilities and Associated Capacities, Adopting Requirements for M-RETS Participation, and Requiring Report Iowa Utilities Board Order, Docket AEP-07-1, Nov. 2007.

[12] Federal Policy, American Wind Energy Association [Online]. Avail- able: http://www.awea.org/issues/federal_policy/index.cfm

[13] C.-J. Yang, E. Williams, and J. Monast, Wind Power: Barriers and Policy Solutions Duke University, Nov. 2008.

[14] Current State of Intraday Markets in Europe ETSO, May 2007. [15] Nord Pool Spot, The Power Market—How Does It Work [Online].

Available: http://www.nordpoolspot.com/How-does-it-work [16] NordPoolSpot IntroducesNegativePricesonElbas,andReducesGate

ClosureonElbas inGermanyPieceofMarketNewsNo.11/2011,Feb. 2011.

[17] H. Holttinen, P. Meibom, A. Orths, F. Hulle, B. Lange, M. O’Malley, J. Pierik, B. Ummels, J. O. Tande, A. Estanqueiro, M. Matos, E. Gómez, L. Söder, G. Strbac, A. Shakoor, J. Ricardo, J. C. Smith, M. Milligan, and E. Ela, Design and Operation of Power Systems With Large Amounts of Wind Power Final Report, IEA WIND Task 25, Phase One 2006–2008, Aug. 2008.

[18] C.Klessmann,C.Nabe, andK.Burges, “Pros andconsof exposing re- newables to electricity market risks—A comparison of the market in- tegration approaches in Germany, Spain, and the UK,” Energy Policy, vol. 36, pp. 3646–3661, 2008.

[19] Act on Granting Priority to Renewable Energy Sources (Renewable Energy Sources Act—EEG), Version Applicable as at 1 January 2012.

[20] Association of Electricity Producers, Electricity Market [Online]. Available: http://www.aepuk.com/about-electricity/electricity-market

[21] SEMO, Single Electricity Market Operator [Online]. Available: http:// www.sem-o.com

[22] ERCOT,TexasNodalMarket ImplementationArchive[Online].Avail- able: http://nodal.ercot.com/index.html

[23] MISO, Market Information [Online]. Available: https://www.mid- westiso.org/MarketsOperations/MarketInformation/Pages/Market- Information.aspx

[24] CAISO,Market Processes [Online].Available: http://www.caiso.com/ market/Pages/MarketProcesses.aspx

[25] J. Rivier Abbad, “Electricity market participation of wind farms: The success story of the Spanish pragmatism,” Energy Policy, vol. 38, pp. 3174–3179, 2010.

[26] F. V. Hulle, N. Fichaux, A.-F. Sinner, P. E. Morthorst, J. Munksgaard, S. Ray, C. Kjaer, J. Wilkes, P. Wilczek, G. Rodrigues, and A. Ara- pogianni, Powering Europe: Wind energy and the electricity grid, A Report by the European Wind Energy Association, Nov. 2010.

818 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

[27] C. Hiroux and M. Saguan, “Large-scale wind power in European elec- tricity markets: Time for revisiting support schemes and market de- signs?,” Energy Policy, vol. 38, pp. 3135–3145, 2010.

[28] A. Helander, H. Holttinen, and J. Paatero, “Impact of wind power on the power system imbalances in Finland,” in Proc. 7th Int. Workshop onLarge-ScaleIntegrationofWindPower intoPowerSystemsMadrid, España, 2008.

[29] NERC, Accommodating High Levels of Variable Generation Apr. 2009, Special Report.

[30] T. Ackermann, G. Ancell, L. D. Borup, P. B. Eriksen, B. Ernst, F. Groome, M. Lange, C. Mohrlen, A. G. Orths, J. O’Sullivan, and M. de la Torre, “Where the wind blows,” IEEE Power Energy Mag., vol. 7, no. 6, pp. 65–75, Nov./Dec. 2009.

Néstor Aparicio (S’06–M’12) received the M.Sc. degree fromtheUniversity JaumeIofCastelló (UJI), Castelló de la Plana, Spain, in 2002, and the Ph.D. degree from Universidad Politécnica de Valencia, Spain, in 2011. He is an Assistant Professor of the Electrical

Engineering Area at Universitat Jaume I, with re- search interests in thegrid integrationofwind-power generators. For 6 months, he visited the Institute of Energy Technology of Aalborg, Denmark and the Centre for Energy and Environmental Markets

(CEEM), Sydney, Australia in 2006 and 2008, respectively.

Iain MacGill (M’10) is an Associate Professor in the School of Electrical Engineering and Telecom- munications at the University of New South Wales, Sydney, Australia, and Joint Director (Engineering) for the University’s Centre for Energy and Environ- mental Markets (CEEM). His teaching and research interests at UNSW include electricity industry restructuring and the Australian National Electricity Market, sustainable energy technologies, renewable energy integration into power systems, and energy policy.

Juan Rivier Abbad (M’99) received the Electronic Engineering degree from the Universidad Pontificia Comillas,Madrid, in1992,andthePh.D.degreefrom the same University in 1999. He joined the Instituto de Investigación Tec-

nológica (IIT) in 1992 as a research fellow and the Electrical Department of the Engineering School (ICAI) in 1999 as an assistant professor, both at the Universidad Pontificia Comillas. He was a Visiting Research Fellow at the Centre for Energy and Environmental Markets (CEEM) of the University

of New South Wales, Australia, during the academic year 2005/06. He is currently the Energy Management Responsible at Iberdrola Renovables. He has experience in industry joint research projects in the field of electric energy systems in collaboration with international and Spanish utilities, and with energy regulatory commissions. His areas of interest are regulation, operation, and integrationof renewableenergysourcesgenerators, andelectricitymarkets.

Hector Beltran received the M.Sc. degree in indus- trialengineering, in2004, fromtheUniversitatJaume I (UJI),Castellóde laPlana,Spain, and thePh.D.de- gree inelectricalengineering, in2011, fromtheTech- nicalUniversityofCatalonia (UPC),Terrassa,Spain. During 2003, he worked at the European Centre

forNuclearResearch (CERN),Geneva,Switzerland. From 2004 to 2006, he worked as a Researcher at theElectronicandEnergyDepartmentsof theEnergy Technologycal Institute (ITE),València,Spain.Since 2006, he is an Assistant Professor in the Electrical

Engineering Area at UJI. Meanwhile, he visited the Institute of Energy Tech- nology, Aalborg University, Denmark (for 6 months), and the Renewable Ener- gies Electric Systems Research Group at the Technical University of Catalonia (UPC),Spain(for9months).Hiscurrent researchinterests includemassivepho- tovoltaic integration into the grid, energy-storage systems, and microgrids.

1

Effects of increasing wind power penetration on the physical operation

of large electricity market systems Bernd Klöckl, Member, IEEE, and Pierre Pinson

Abstract—This contribution describes indirect coupling effects between wind power infeed and physical operation of power market systems by means of qualitative hypotheses, backed by suited exploratory data analyses. As an example, the case of a central European TSO located in the vicinity of a control block with high wind penetration is demonstrated. It shows, based on established methods of computational statistics, considerable nonlinear effects on cross border power flows and transmission system flows of that control block, conditional on increasing wind power penetration in an interconnected market system. The observed effects are theoretically explained through the influence of the wind infeed on the behaviour of market participants and attributed to indirect coupling between wind power and conventional generation in adjacent control blocks.

Index Terms—Wind power, energy market, cross border flows, principal component analysis.

I. INTRODUCTION

T HE rapidly increasing share of wind generation in theEuropean energy markets, together with further progress in their liberalization has led to a number of direct and indirect

effects of wind power injection on flow patterns in the trans-

mission systems and has increased the operational challenges

for TSOs. The developments throughout the last years have

shown that increasing the share of wind generation beyond

certain levels has created the need for further investigations

on the following issues:

1) connection issues and grid codes (technical)

2) high concentrations of volatile infeeds in certain areas,

e.g. close to windy coast lines, leading to the need for

infrastructure reinforcements, especially in the transmis-

sion system (technical)

3) implications of day-ahead wind forecasts on the whole-

sale electricity prices in a given market area (economic)

4) coupling of these implications to adjacent markets and

TSOs (economic)

5) increasing wide-scale influence of wind generation on

the behaviour of conventional generation (economic)

6) thus increasing operational challenges for TSOs beyond

the effect of the wind infeed itself (techno-economic).

The first two points above are currently being extensively

treated by planning and operation staff of DSOs and TSOs,

and the last point is experienced more and more by the grid

B. Klöckl is with the Market Management Department of Verbund Austrian Power Grid, Vienna, Austria, EU (e-mail: [email protected])

P. Pinson is with the Technical University of Denmark, DTU Informatics, Kgs. Lyngby, Denmark, EU (e-mail: [email protected])

operation staff. In this paper, the authors elaborate on the

points 3 to 5, which is a simplified functional chain creating a

feedback loop between technical and economic issues of wind

generation.

The remainder of the paper is structured in the following

way: First, a general overview of the EU transmission system

with respect to the market operation and a glimpse on the

most obvious effects of wind infeed is given. Then, a suited

statistical analysis method for the problem set is derived. This

analysis is exemplified on the measurements taken from the

system of a central European TSO being indirectly affected

by large wind power penetrations in adjacent markets. The

conclusions attempt to relate the observed effects to future

investigation needs.

II. EFFECTS OF WIND POWER ON MARKET PRICES

A. The EU market and transmission system

Fig. 1. Overview of the ETSO member states and the related EU market zones. The geographical distance between northern Germany and the APG block as indicated is approximately 750 km.

The European transmission system operators are organized

in ETSO [1], which is an association that covers five different

synchronous zones. The largest zone of the EU transmission

network is the UCTE grid, that links all national transmission

systems of continental Europe except for the Baltic states

(Fig. 1). It is one of the largest synchronous systems in the

world with an installed generation capacity of approximately

650 GW [2]. For the time being, the different national trans-

mission networks represent market zones that are technically

© CIGRE2009 978-2-85873-080-3

2

connected by sets of tie lines and commercially linked by

allocation mechanisms for the cross-border transmission ca-

pacities.

B. Observed interaction of wind generation and spot prices

Fig. 2. Potential shift of the market settlement price on the aggregated merit order curve of a price zone in dependence on wind generation. Quantiles of the merit order curve (e.g. a 90% quantile) are indicated as an illustration of the uncertainty of the market settlement prices as a consequence of market imperfections and other factors.

Recently, the effect of wind power on market prices has

been discussed for the cases of market areas with high wind

penetration. Generally, there seems to be an agreement that

wind generation lowers the spot market prices, while the

mechanisms behind are not entirely clear. For a first general

treatment of the effect, see e.g. [3]. In [4], the reduction of

price in dependence on wind generation is described. This

does not come as a big surprise since wind generation in the

German system is prioritized in the dispatch and can thus

be regarded as a negative load. In [5], this is interpreted

as an effect on the activation of the merit order curves of

the GENCOs, meaning that in the presence of wind there

is increased probability that expensive power plants will not

settle the spot prices. This effect is then opposed to the costs

of the renewable energy support scheme1. See Fig. 2 for an

graphical illustration of the mechanism. Jóhnsson [6] provides

detailed modeling recommendations for the price reduction

effect applied to the Danish case and states that the decisive

variable for the reduction effect is clearly the wind power

prediction, rather than the actual production at the time of

delivery.

In addition to the findings cited above, it can be shown that

there is a second effect of high wind generation on the market

prices which has not yet been discussed in the literature. One

would assume that, especially for market zones with priority

wind dispatch, the variable to look at is simply a fictitious

expected system load,

Lf ic = L − P̂w, (1) 1However, the reduction of energy price for the end consumers must not

be confused with the calculation of the total socio-economic benefit, which has to take into account also the financial impact on the GENCOs, the wind turbine manufacturers etc.

where L is the system load (the consumption within the market area), and P̂w is the predicted wind generation. Fig. 3 shows the German EEX spot prices for 2006-07 in a logarithmic scale

plotted against Lf ic. It can be observed that for equal Lf ic, the mean prices at high wind generation are still slightly lower

and that price spikes virtually do not occur in periods of high

forecasted wind penetration

r̂w = P̂w/L. (2)

Fig. 3. EEX market prices for 2006 and 2007 in dependence on different levels of Lf ic (Sources: [2], [7], [8], [9], [10]), plotted for events with a wind penetration below 7.8% (red) and above 7.8% (green), where 7.8% is the mean of the wind penetration in the period. The logarithmic scale for the spot price is depicted on the l.h.s., while the plot on the r.h.s. shows the mean values in a linear scale. It can be observed that for the region in which the resulting loads to be covered by the market are equal, the price is still always lower for high r̂w . This might be caused by price expectations of the traders or by the fact that not the entire volume of energy consumed in Germany is traded at the power exchange.

There is little doubt that this general price damping effect of

predicted wind infeed has an influence not only on the affected

market zone itself, but also on adjacent interconnected areas

with correlated market prices. In the following, we would like

to explain how this can be detected from observed data.

III. METHODOLOGY: CONDITIONAL MULTIVARIATE DATA

ANALYSIS

Analysing the operation of a complex system may translate

to simultaneously studying the behaviour of a large number

of possibly redundant variables in a multivariate data analysis

framework. At a given time t the measured values for this set of m variables are gathered in a single vector Xt. For the example of the present study, these variables may be

cross-border flows or flows on transmission systems. Owing

to the instantaneous nature of electricity transmission, flow

data recorded at different points in the horizontal network

may often cloud the true underlying mechanisms by containing

redundant information and noise components that arise from

mechanisms other than the one that is subject to investigation.

A possibility to alleviate this problem is to employ Principal

Component Analysis (PCA) for dimension reduction. This will

3

be described in a first part below. One can then work in a

reduced basis defined by principal components, and develop

conditional parametric models for capturing the nonlinear

effects of the influential variables of interest, e.g. forecasted

wind power penetration, on the flows.

A. Dimension reduction with Principal Component Analysis (PCA)

PCA is a classical method in multivariate data analysis,

which allows one to reduce the dimension of the problem at

hand, and to potentially work in a reduced orthonormal basis

defined by the set of (uncorrelated) principal components. For

a nice introduction to multivariate data analysis and PCA, the

reader is referred to [11], while more extensive mathematical

developments may be found in e.g. [12]. Consider a number

N of measured flow values Xt (being of dimension m), and define X̃t the centered and standardized version of Xt. This simply means that for each dimension of X̃t one has

X̃t,j = τj (Xt,j ) = Xt,j − X̄j

σj , j = 1, . . . , m (3)

where X̄j and σj are the mean and standard deviation of the jth flow variable. We will denote by τ the simultaneous application of the τj transformations to all components of Xt.

Finding the principal components for the dataset considered

may be performed by diagonalizing the covariance matrix of

the data, given as

R = 1 N

N∑

i=1

X̃tX̃ � t (4)

By arranging the eigenvalues in decreasing order and identi-

fying the corresponding eigenvectors, one obtains the set of

principal components. Using the average eigenvalue method

[12, pp. 348], the set of retained principal components are

those for which the related eigenvalue is larger than the mean

eigenvalue of R. By writing Yi (i = 1, . . . , n, n < m) these principal components, one then obtains an orthonormal basis in

which X̃t can be written as linear combination of the principal components,

X̃t = n∑

i=1

αiYi + �t, ∀t (5)

plus �t which is a m-dimensional centered noise of finite variance. By comparing the eigenvalues corresponding to the

retained principal components to the trace of R (i.e. the sum of all eigenvalues), one can determine the degree of

explanation of the variance in the data, giving a hint on the

information content of the remaining noise. Note that the

principal components Yi can be seen as orthogonal modes explaining the interrelated variations of the flows considered,

while

P = [Y1 Y2 . . . Yn] (6)

can be seen as the projection matrix allowing to project

standardized flow values X̃t in the basis defined by the principal components.

B. Conditional parametric models for local smoothing

The model in Eq. (5) permits one to express the flows as

a linear combination of the principal components Yi. Such models can be extended in order to account for the potential

nonlinear effects of influential variables on the flows. Denote

by ut the values of these influential variables at time t. They may include forecast wind power penetration, fictitious load or

the spot price on the EEX market for instance. The dimension

of ut should be kept low (say, lower than 3) owing to the so- called curse of dimensionality. The model of Eq. (5) is then

extended to

X̃t = n∑

i=1

αi(ut)Yi + �t, ∀t (7)

where the αi coefficients are not constant anymore, but instead coefficient functions of the set of influential variables ut. �t is still a m-dimensional centered noise of finite variance.

We do not describe here the detail of the method for

estimating the coefficient functions. In general, the challenge

is to define a fitting procedure for the coefficients αi(ut) that allows to consistently eliminate �t in Eq. (7) (i.e. to regard it as noise component). The basis for their estimation can be

found in e.g. [13]. No assumption is made about the shape of

the coefficient functions, except that they are continuous and

suffiently smooth for being locally approximated. The method

for their estimation consists of approximating them locally

at a number of fitting points with first order polynomials,

and of using weighted least squares for determining the

polynomial coefficients. Different variants of the method for

their estimation can be found in e.g. [14], [15].

C. Identifying trend surfaces

There may be different ways of using the estimated coef-

ficient functions in Eq. (7) for analysing the impact of the

defined influential variables on the flows. One can in a first

stage analyse the estimated α̂i functions themselves in order to see how influential variables act on the contribution of the

various identified modes to the observed flows. Alternatively,

it may be easier to go back to the original flow variables and

to show what is the mean effect of the influential variables

in the variations of the various flows considered. It is this

alternative that will be preferred in the following. Indeed, by

simply discarding the noise term in Eq. (7), projecting the α̂i functions back to the basis in which X̃ is defined, and using the inverse τ transformation for getting back to the original X variables, one obtains

X̂(u) = τ −1 (Pα̂i(u)) (8)

which defines the mean flows as a function of the influential

variables u. Variations in such mean flows as a function of u can be seen as trends induced by u, which can be for instance a trend induced by forecast wind power penetration. Examples

of such trend surfaces will be given and discussed below.

4

IV. METHODOLOGY APPLICATION EXAMPLE

A. Influential variables in the analysis

For this study, a closer look is taken at the control block

of Austria, operated by the TSO APG (Fig. 1). The installed

generation capacity in Austria is more than 19 GW (12 GW

in hydro power units), while the maximum load is less than

10 GW [2]. In spite of this excess of installed generation

capacity, the block has shifted its overall characteristics from

export to import throughout the last years, which can be caused

by a number of factors. The block is physically linked to six

different control blocks and a total of nine control zones.

The complexity of such structures is demanding a concise

top-down approach based on recorded data. Effects caused

by the following influential variables were subject to detailed

investigations:

1) fictitious load (Eq. (1)) in Germany, and

2) forecasted wind power penetration in Germany (Eq. (2)),

both together generating an indirect effect on generation and

trading within the APG block via market-related effects. The

trends of cross-border flows and APG system flows were

subsequently analysed for 2006-07 data according to the

methodology outlined in Sec. III above. The vector ut as introduced in Eq. (7) then includes the measured values for

these 2 influential variables at time t. In parallel, the vector Xt of flow values measured at time t may either relate to the overall block balance (being one-dimensional in this case,

thus not needing the PCA step of the methodology introduced

above), or gathering the set of cross-border flows (m = 6), or finally the flows on transmission lines (m = 23).

B. Flow patterns as a result of wind power

Fig. 4. Trend of the block balance of APG in dependence on the wind power penetration in Germany.

1) Control block balance: The control block of APG is closely linked to the control block of Germany, due to high

transmission capacities and mutual benefits in the generation

mixes the two block are able to share. In particular, the pumped

hydro facilities within the APG area serve as buffer for low

Fig. 5. Trend of the block balance of APG in dependence on the wind power penetration and the fictitious load in Germany. Due to computational reasons, the three-dimensional fitting procedure cannot be applied with suffcient numerical reliability up to the full measured wind power penetration of 40%.

price energy injected into the German block. The question

is if a general tendency for a relation between the wind

penetration in Germany and the export/import balance of APG

can be found. Since there are many different simultaneously

relevant influential variables, the method described in Sec. III

has been used for the identification of the influence of the

wind power penetration alone. The result is depicted in Fig. 4

and shows an interesting feature: The general trend indicates

that for zero wind penetration in Germany, there is even a

slight export to be expected from the APG block. For higher

penetrations, the sign turns to negative, that is, the consumers

and pumped hydro plants clearly start to import “green”

energy. Disintegrating this information to the influence of the

wind power penetration and the ficititious load in Germany

separately results in the trend surface plot shown in Fig. 5.

It shows the trends of the export/import balance of APG in

dependence on the wind power penetration and the fictitious

load in Germany at the same time.

2) Physical cross border flows: The cross-border flow trends of the APG block have been identified following the

methodology outlined in Sec. III. The dimension reduction

part of the methodology has led to the identification of 2

modes explaining 60% of the variations in the data. The

resulting trend computations are shown in Figs. 6, 7 and 8.

The analysis indicates that for high wind penetration and low

load in Germany, the northern flowgates of APG (Germany and

Czech Republic) are importing considerably more energy than

in periods of low wind penetration and high load in Germany.

For the export flow at a southern flowgate to Switzerland (and

then further to Italy), the opposite is the case, which is a clear

indication of wind energy transit through the APG block.

3) Unintended cross border flows: The unintended cross- border flow at a flowgate is defined as the difference between

the physical flow and the scheduled flow. It turns out that the

level of wind power penetration influences the probability of

unintended exchanges on APG’s borders. Again, the dimension

5

Fig. 6. Trend of the physical cross-border flow at the border between APG and Germany dependent on the wind power penetration and the fictitious load in Germany.

Fig. 7. Trend of the physical cross-border flow at the border between APG and Czech Republic dependent on the wind power penetration and the fictitious load in Germany.

reduction part of the methodology has led to the identification

of 2 modes, which explain here 67% of the variance in the

data. As an example, Fig. 9 shows the trends at the German

border. There is a nonlinear dependence on the fictitiuos

load, showing a peak at a certain value, and a roughly linear

and clearly positive dependence on the German wind power

penetration. Here, it should be noted that a part of such effects

comes from the zonal market model and is influenced by

the impedance relations in the horizontal network. That is,

the overall unintended exchange of the APG block is widely

unaffected by the wind power penetration.

4) Transmission line flows: The 4 identified modes (ex- plaining 75% of the variance of the original data) of some

important 400-kV-systems have been analysed in order to

distinguish between more and less volatile flow patterns. In

Fig. 10, these modes are plotted for 23 systems. Following the

Fig. 8. Trend of the physical cross-border flow at the border between APG and Switzerland dependent on the wind power penetration and the fictitious load in Germany.

Fig. 9. Trend of the unintended exchange at the border between APG and Germany dependent on the wind power penetration and the fictitious load in Germany.

introduction of PCA in Section III-A, most of the variations

of the flows on the 400-kV-systems can be explained and

expressed as a linear combination of these 4 modes. Visual

inspection can already tell which systems have more flow

variations. It can be seen that e.g. system No. 15 shows distin-

guished volatile behaviour, since exhibiting larger magnitude

of variations in the various modes. Fig. 11 reveals that the

flow on this line shows a clear nonlinear dependence on the

wind power penetration and the fictitious load in Germany.

V. CONCLUSIONS

The market-related indirect influence of high local wind

penetration on physical variables in other locations of large

transmission systems has been demonstrated for an EU control

block. The theoretically suspected effect has been separated

from other influences by means of a classical multivariate data

6

Fig. 10. Modes of 23 400-kV-systems in the APG block identified by PCA as described in Sec. III.

Fig. 11. Trend of the flow on one selected APG system (No.15) dependent on the wind power penetration and the fictitious load in Germany.

analysis method. Since the future development of renewables

in the EU is highly ambitious, there will be the need for a

projection of the observed nonlinearities to future renewables

penetrations. Top-down approaches based on measured data

as the one outlined in this paper seem to be a practical and

reliable way to approach this problem. In the light of the

demonstrated statistical evidence for still relatively small wind

power penetrations incorporated into the market via a simple

mechanism, future research activities should increasingly fo-

cus on smart market design for renewables and other new

energy technologies.

REFERENCES

[1] European Transmission System Operators (ETSO). information online. http://www.etso-net.org, 2009.

[2] Union for the Co-ordination of Transmission of Electricity (UCTE). values online. http://www.ucte.org, 2009.

[3] T. Ackermann (Ed.) and P.E.Mothorst. Wind Power in Power Systems, chapter Economic Aspects of Wind Power in Power Systems, page 401 ff. Wiley, 2005.

[4] J. Neubarth, O. Woll, C. Weber, and M. Gerecht. Beeinflussung der Spotmarktpreise durch Windstromerzeugung. Energiewirtschaftliche Tagesfragen, 56:42–45, 2006.

[5] F. Sensfuß and M. Ragwitz. Analyse des Preiseffekts der Stromerzeu- gung aus erneuerbaren Energien auf die Börsenpreise im deutschen Stromhandel. Technical report, Fraunhofer ISI, 2007.

[6] T. Jónsson. Forecasting of Electricity Prices Accounting for Wind Power Predictions. Master’s thesis, Technical University of Denmark, Lyngby, 2008.

[7] European Energy Exchange. values online. http://www.eex.com, 2009. [8] E-ON Netz. values online. http://www.eon-netz.com, 2009. [9] RWE Transportnetz Strom. values online.

http://www.rwetransportnetzstrom.com, 2009. [10] Vattenfall Europe Transmission. values online. http://www.vattenfall.de,

2009. [11] J. Lattin, J.D. Carroll, and P.E. Green. Analyzing Multivariate Data.

Duxbury Applied Series, 2003. [12] A.C. Rencher. Multivariate Statistical Inference and Applications. Wiley

Series in Probability and Statistics, 1998. [13] W. Cleveland and S. Devlin. Locally Weighted Regression: An Approach

to Regression Analysis by Local Fitting. Journal of the American Statistical Association, 83(403):596–610, 1988.

[14] H.Aa. Nielsen, T.S. Nielsen, A.K. Joensen, H. Madsen, and J. Holst. Tracking Time-varying Coefficient Functions. Int. J. Adapt. Control, 14:813–828, 2000.

[15] P. Pinson, H.Aa. Nielsen, H. Madsen, and T.S. Nielsen. Local Linear Regression with Adaptive Orthogonal Fitting for the Wind Power Application. Stat. Comput., 18:59–71, 2008.

Bernd Klöckl (M’02) received the M.Sc. degree in Electrical Power Engineering from Graz University of Technology, Austria, in 2001 and the Ph.D. degree from ETH Zurich, Switzerland, in 2007, where he was research associate and lecturer for renewables from 2002 to 2006. From 2006 to 2007, he headed the grid section of the Austrian Association of Elec- tricity Companies, Vienna, Austria. Since 2007, he is with the national TSO APG, responsible for research in market models and cross-border tariffication in ETSO. Dr. Klöckl is member of IEEE and CIGRÉ.

Pierre Pinson received the M.Sc. degree in Applied Mathematics from the National Institute for Applied Sciences (INSA Toulouse, France) in 2002 and the Ph.D. degree in Energetic from Ecole des Mines de Paris in 2006. He is currently with the Informat- ics and Mathematical Modeling department of the Technical University of Denmark as an Associate Professor. His research interests include among oth- ers forecasting, uncertainty estimation, optimization under uncertainty, decision sciences, and renewable energies.

1   2   3   4   5   6   7   8   9   10   11   12   13   14   15   16   17   18   19   20   21   22   23   24   25   26   27   28   29   30   31   32   33   34   35   36   37   38   39   40   41   42   43   44   45   46   47   48   49   50   51   52   53   54   55   56   57   60   61   62   63   64   65  

Electricity Price Variations in a Power System with Wind Energy Penetration

Avinash D.

M. Tech. Student, Electrical Engineering Department National Institute of Technology, Kurukshetra

Haryana, India [email protected]

Shelly Vadhera Assoc. Prof., Electrical Engineering Department National Institute of Technology, Kurukshetra

Haryana, India [email protected]

Abstract — This paper aims to give the variations of electricity

price by considering the effect of wind energy integrated into the system and also the system variations such as load demand, outages and congestion. The electricity market is cleared by using a two stage stochastic optimization model in which first stage represents the scheduled case and the second stage represents the deviations from the scheduled case caused by uncertainty in the production of wind energy, this stage is considered as the balancing market operation. IEEE 24 bus test system is considered in different system conditions with 26.3% of wind energy penetration to study the price variations.

Keywords — Congestion, Electricity prices, Stochastic optimization, Wind energy.

I. INTRODUCTION Wind energy is been considered as the best alternative

compared with the other renewable energy sources, according to Global Wind Energy Council, (GWEC) [1] 2014 was a record year for wind industry crossing the 50 GW of global annual installation, with this the cumulative installed capacity of wind energy at the end of 2014 around the world is about 369,597 MW [1].

This shows the increased integration of wind energy into the power system network which affects the various parameters like electricity prices, congestion and reliability. The electricity market considered in this paper is based on pool operation where the market is cleared in two stages one in day ahead stage, that is 24 hours before the actual operation and the other in balancing stage where the adjustments to the scheduled case are considered. Balancing market operations are carried at the actual time of dispatch which varies according to the wind production level and other uncertainties. A mathematical model is presented in [2] to co optimize the energy and reserve, at a given load demand to get the prices in two stages and to schedule the reserves according to the wind variations.

In [3] a literature review is presented for the European Wind Energy Association (EWEA) to show the decrease of spot prices because of wind power production, which is verified by a regression analysis. The correlation between the two adjacent wind farms and how one wind farm affects the other is considered in [4] using a stochastic programming optimization technique. The wind integration also creates the

problem in planning and scheduling the units, in recent years combined Energy Storage System with wind energy as in [5] are proposed to decrease the effect of uncertainty of wind production, several other methods such as demand response in which demand responds to the variations of prices, especially in peak hours and in wind power generation scenarios are developed [6] and in [7] game theory approach is used for planning problems. With respect to smart grid applications there are some new approaches developed incorporating orthogonal frequency division multiple access (OFDMA) in [8 – 10]. Pricing based on stable matching algorithm and message-passing algorithm for data redistribution market in wireless networks is considered in [11]. In this paper stochastic programming is used to schedule the units, since prices are dual variables of the power balance constraints they can be obtained at both the day ahead and balancing stages.

Section II of the paper describes the modeling of stochastic programming technique which is used to clear the electricity market, whereas section III gives the results for IEEE 24 bus test system and section IV concludes the paper.

II. ELECTRICITY MARKET MODEL Here, pool operation of the market is considered as one

sided bidding in which sellers submit their bids to the pool operator, based on the load demand, the generation and reserves are scheduled to obtain the minimum cost of operation. DC load flow method is used to simplify the model and hence the model can be solved by Linear Programming (LP) techniques

A. Motivation Stochastic programming can be used for the decision

making problems in which uncertainty is present [12], with the integration of wind energy the uncertainty in power generation is increased along with the other uncertainties such as load variations and electricity prices. The uncertain wind power data is given as a set of possible scenarios, each having certain probability at a given point of time, the output consists of schedule of the units and respective prices for each and every scenario at all the considered time periods.

978-1-4673-6540-6/15/$31.00 ©2015 IEEE

IEEE INDICON 2015 1570168503

1

B. Algorithm • Scheduled wind generation is included in the day

ahead market stage where as the deviations of Wind generation are considered as low wind and high wind scenarios in balancing market stage, for each time period

• Wind generation is modeled as the negative load demand

• Reserve bidding is included simultaneously with the generation scheduling trading according to [13]

• A day schedule is divided into 24 time periods each indicating 1 hour of operation, thus the solution gives a dispatch schedule for each hour for the upcoming day

C. Formulation Minimize

( )⎥ ⎦

⎤ ⎢ ⎣

⎡ −∏+

++

∑∑

∑

∈∈

∈

Kk

D kst

D kt

U kst

U kt

Ss s

D kt

RD kt

U kt

RU kt

Kk ktkt

rCrC

RCRCPC

TtLV Ll

shed lst

LOL l

Ss s ∈∀⎥

⎦

⎤ ⎢ ⎣

⎡ ∏+ ∑∑

∈∈

: (1)

Subject to

( ) 000 =−−−+ ∑∑∑∑ Φ∈Φ∈Φ∈Φ∈

mtnt m

nm l

nt q

sh qt

k kt

M n

L n

Q n

K n

bLWP δδ

: , , (2) ( ) ( )∑∑∑

Φ∈Φ∈Φ∈

−−++− Q n

L n

K n q

spill qst

sh qtqst

l

shed lst

k

D kst

U kst WWWLrr

( ) 000 =+−−+ ∑ Φ∈

mstmtnstnt m

nm M n

b δδδδ

: , , , (3) tkPP ktkt ∀∀≤ ,,

max (4)

tskrrP Dkst U kstkt ∀∀∀≥−+ ,,,0 (5)

tskPrrP kt D

kst U kstkt ∀∀∀≤−+ ,,,

max (6)

( ) ( ) tmnCb nmmtntnm ∀Λ∈∀≤− ,,,max00 δδ (7) ( ) ( ) tsmnCb nmmstnstnm ∀∀Λ∈∀≤− ,,,,maxδδ (8)

tkRR uk U kt ∀∀≤≤ ,,0

max (9)

tkRR dk D kt ∀∀≤≤ ,,0

max (10)

tskRr Ukt U kst ∀∀∀≤ ,,, (11)

tskRr Dkt D

kst ∀∀∀≤ ,,, (12)

tslLL l shed lst ∀∀∀≤ ,,, (13)

tsqWW qst spill

qst ∀∀∀≤ ,,, (14) tkPP ktkt ∀∀≥ ,,

min (15)

tskrr Dkst U kst ∀∀∀≥ ,,,0, (16)

tsqWtqW spillqst sh

qt ∀∀∀≥∀∀≥ ,,,0;,,0 (17) tslLshedlst ∀∀∀≥ ,,,0 (18)

0 ntδ free nstn δ;, ∀∀ free tsn ∀∀∀ ,, (19)

Where equation (1) is the objective function representing the total cost of operation which is to be minimized subject to the constraints given by equations (2) to (19), and are the set and index of generator units respectively. , , and , are the production cost, upward/downward reserve cost and the cost for increasing/decreasing the generation in balancing stage for unit at time period . The set of all time periods is indicated by , equation (2) is a power balance equality constraint for which the day ahead prices are the dual variables represented by where , both used as the index of nodes, , , , represents the power generation, scheduled wind power, load demand and the susceptance respectively, and are used to indicate the set and index of wind generators respectively. , , , are up/down reserves and increasing/decreasing power levels in balancing stage respectively, , indicate the value of lost load, load shed in scenario at time , and phase angles of the node in day ahead stage respectively. , are the actual wind power generated and the wind power spilled because of the line flow constraints in scenario at time . Equation (3) is the power balance constraint equation in balancing stage in which price is a dual variable. Equations (4) to (18) are inequality constraints in which , , , , , are the limits of generation, line flows, up/down reserves and the load present at a load point respectively. Φ , Φ , Φ are the sets which relate the set of nodes to generators, load points, and wind generators where as Φ gives the set of adjacent lines. Π represents the probability of wind generation scenario .

D. Assumptions • Wind energy is considered as the freely available

energy hence no cost is associated with it • Inelastic loads are considered • Since DC power flow is used network losses and

reactive power flows are neglected • Fuel cost of the units is assumed to be a piece wise

linear curve

III. RESULTS IEEE 24 bus reliability test system [14] is considered in

which there are 12 generating units present, which supply a load of 2850 MW through 34 transmission lines. Cost data of the system is taken from [15], according to [2] wind farms are

2

connected at buses 7 and 8 having a capacity of 23.6% of the total demand i.e. 750 MW, thus the two wind farms have a total of 300 wind turbines of 2.5 MW each in which, 7th and 8th nodes have 100 and 200 wind turbines respectively [2]. There is a double line circuit exist between the nodes 15 and 21 carrying a total power of 448.6 MW, for analyzing the effect of congestion on the system prices a single line circuit is considered in place of the double line hence reducing the capacity of line flow to 400 MW. Fig. 1 shows the price variations in day ahead market stage giving the comparison between the normal and the congestion cases, where horizontal

Fig. 1. Price variations in day ahead stage comparing the normal and

congestion cases

Fig. 2. Price variations in low wind scenario comparing the normal and

congestion cases

axis represents the system nodes and the vertical axis represent the price in $/MWh. In normal condition of operation the prices at all nodes are nearly equal but in congestion case, price varies randomly between the nodes 15 and 21 showing a maximum value of 24.917 $/MWh at 15th bus and a minimum value of 5.47 $/MWh at bus 21. Fig. 2 and Fig. 3 shows the price variations for low wind and high wind scenarios, it can be seen that in low wind energy scenario the price is almost constant in normal case but the value is at 20.93 $/MWh which is higher than the day ahead average price (13.32 $/MWh). In congestion case the cost variation is similar to the previous case but the variation of price is reduced between the

Fig. 3. Price variations in high wind scenario comparing the normal and

congestion cases

nodes 5 to 15 indicating the effect of scheduled wind energy in the day ahead stage. The average price for both the normal and congestion cases is reduced drastically in high wind energy scenario because of the availability of wind generation to its maximum capacity (26.3% of total load). The price variations also increased showing a minimum value at buses 7 and 8, since the wind farms are connected at 7th and 8th buses the price varied randomly between the adjacent nodes due to line flow constraints.

In Fig. 4 and Fig. 5 the effect of varying load on nodal prices is shown in normal and congestion cases respectively, where the horizontal axis represents the percentage of the nominal load of the system and the vertical axis represents the cost in $/MWh, in normal case as the load increased the cost of the nodal prices are increased. Except the nodes 7 and 8 where wind farms are connected, remaining all nodes have a similar price pattern as shown in Fig. 4, but for the case of congestion node 15 has a maximum cost where as the cost at node 21 is constant for all the loading conditions as in Fig. 5.

0

5

10

15

20

25

30

1 3 5 7 9 11 13 15 17 19 21 23

P ri

ce (

$/ M

W h

)

Nodes

Normal Congestion

0

5

10

15

20

25

30

1 3 5 7 9 11 13 15 17 19 21 23

P ri

ce (

$/ M

W h

)

Nodes

Normal Congestion

0

2

4

6

8

10

12

14

16

18

20

1 3 5 7 9 11 13 15 17 19 21 23

P ri

ce (

$/ M

W h

)

Nodes

Normal Congestion

3

Fig. 4. Price variations in selected nodes at different loadings under normal

conditions

The above system is solved using the model developed in section II which is implemented in General Algebraic Modeling System (GAMS) as a linear programming problem. The solver used is the CPLEX 12.1.0 under GAMS, the CPLEX’s state of the art dual simplex algorithm is used in solving the model.

Fig. 5. Price variations in selected nodes at different loadings in a congested

case

IV. CONCLUSION This paper gives the variations of electricity prices in the

power system network considered by taking into account the various parameters such as wind generation, load fluctuation, and congestion. Though the wind generation decreases the cost of operation of the system, it increases the variation of price

among the nodes. The same analysis can be extended by considering start up and shut down cost of units and also the cost of wind generation to get more reliable price variations.

REFERENCES [1] [online]. Available: http://www.gwec.net/global-figure [2] J. M. Morales, Antonio J. Conejo, Kai Liu, and Jin Zhong,

“Pricing electricity in pools with wind producers,” IEEE Trans. Power Syst., vol. 27, no. 3, pp. 1366-1376, Aug. 2012.

[3] [online]. Available: http://www.ewea.org [4] J. M. Morales, A. J. Conejo, and J. P. Ruiz, “Simulating

the impact of wind production on locational marginal prices,” IEEE Trans. Power Syst., vol. 26, no. 2, pp. 820 - 828, May 2011.

[5] Maria Dicorato, Giuseppe Forte, Mariagiovanna Pisani and Michele Trovato, “Planning and operating combined wind – storage system in electricity market,” IEEE Trans. Sust. Energ., vol. 3, no. 2, pp. 209 – 217, April 2012.

[6] Cedric De Jonghe, Benjamin F. Hobbs and Ronnie Belmans, “Optimal generation mix with short term demand response and wind penetration,” IEEE Trans. Power Syst., vol. 27, no. 2, pp. 830 – 839, May 2012.

[7] Ting Dai and Wei Qiao, “Trading wind power in a competitive electricity market using stochastic programming and game theory,” IEEE Trans. Sust. Energ., vol. 4, no. 3, pp. 805 – 815, July 2013.

[8] Dexiang Zhan, Haijun Zhang, Zhaoming Lu, Xiangming Wen and Yawen Chen, “Resource allocation for OFDMA two – way relay networks with the smart grid,” Proc. IEEE WPMC, pp. 657 – 662, Sept. 2014.

[9] Wenmin Ma, Haijun Zhang, Wei Zheng and Xiangming Wen, “Differentiated – Pricing based power allocation in dense femtocell networks,” Proc. IEEE WPMC, pp. 599 – 603, Sept. 2012.

[10] Wenmin Ma, Wei Zheng, Haijun Zhang, Xiangming Wen, Zhaoming Lu and Deli Liu, “Pricing based resource allocation in downlink multi – cell OFDMA networks,” Proc. IEEE ICCNT, pp. 260 – 263, Aug. 2012.

[11] Yanyao Shen, Chunxiao Jiang, Tony Q. S. Quek, Haijun Zhang and Yong Ren, “Pricing equilibrium for data redistribution market in wireless networks with matching methodology,” Proc. IEEE ICC, pp. 3051 – 3056, June 2015.

[12] J. M. Morales, A. J. Conejo, Henrik Madsen, Pierre Pinson, and Marco Zugno, Decision making under uncertainty in electricity, operational problems. New York: Springer 2014.

[13] J. M. Morales, A. J. Conejo, Henrik Madsen, Pierre Pinson, and Marco Zugno, Integrating renewables in electricity markets, operational problems. New York: Springer 2014.

0

5

10

15

20

25

60 70 80 90 100 110 120

P ri

ce (

$/ M

W h

)

Load (%)

Node 1 Node 7

Node 15 Node 21

0

5

10

15

20

25

30

35

40

45

60 70 80 90 100 110 120

P ri

ce (

$/ M

W h

)

Load (%)

Node 1 Node 7

Node 15 Node 21

4

[14] Reliability Test System Task Force, “The IEEE reliability test system – 1996,” IEEE Trans. Power Syst., vol. 14, no. 3, pp. 1010-1020, Aug. 1999.

[15] [online]. Available: http://pierrepinson.com

APPENDIX In this model, cost for increasing/decreasing the generation in balancing stage is taken as the production cost. The other parameters used in the model are given in Table I.

TABLE I. COST PARAMETERS

Units (k)

($/MWh) ($/MW) ($/MW)

1 13.32 15 14

2 13.32 15 14

3 20.70 10 9

4 20.93 8 7

5 26.11 7 5

6 10.52 16 14

7 10.52 16 14

8 5.47 0 0

9 5.47 0 0

10 0 0 0

11 10.52 17 16

12 10.89 16 14

5

1

Abstract—Wind generation is playing an increasingly

significant role in many electricity industries around the world. It has very different operational characteristics from existing generation and the integration challenge is to facilitate wind in achieving its maximum energy, environmental and wider societal value. Maximizing energy value is a particular challenge for electricity arrangements as penetrations increase. This paper summarises some of the key design features of the Australian National Electricity Market (NEM) relevant to wind integration, assesses its performance to date in facilitating appropriate wind deployment, and outlines possible changes to support much higher wind penetrations in the future. It suggests that the NEM has reasonably effectively and efficiently managed regionally significant wind penetrations to date. Wind is now having significant impacts on market outcomes in high penetration regions. In particular, periods of high wind output are associated with lower wholesale market prices and wind is receiving lower revenue ($/MWh) than other generation types. This reflects the lower energy value of a generation source relying on a non-storage energy source. Projected significant increases in wind penetration due to government renewable energy targets for 2020 will continue to test the adequacy of market arrangements and likely require ongoing market design changes. These changes need to be seen in the wider context of NEM challenges in transitioning towards a low-carbon future.

Index Terms— Australia, Electricity market, Wind integration, market design

I. INTRODUCTION Ind generation is playing an increasingly significant role in many electricity industries around the world and

Australia is no exception. Over 2GW of wind generation has now been deployed and it is expected that current government policy will drive considerable further investment over the coming decade.

In its broadest sense, the challenge for wind integration within electricity industries is to facilitate wind energy in achieving its maximum energy, environmental and wider

I. F. MacGill is with the Centre for Energy and Environmental Markets

and School of Electrical Engineering and Telecommunications at the University of New South Wales, Sydney, Australia (email: [email protected]).

societal value. Maximizing energy value is a particular challenge as wind penetrations increase and has to be addressed in the context of the electricity industry as a whole. Wind’s energy value depends in part on the investment and operational costs of particular wind farms. However, its value is also determined by the impact it has on the benefits and costs of other power system participants. These costs and benefits have temporal and locational variability and uncertainty that emerge from the coordinated behaviour of all loads and generation on a shared network

Electricity industries, of course, face many challenges other than wind energy integration, which should therefore be considered in a broader context that addresses issues including the security and efficiency of delivering energy services to end users, and appropriately supporting other opportunities for meeting our growing energy security and greenhouse challenges.

Wind energy, however, represents the first highly non- conventional generation to reach significant penetrations in large power systems, and is therefore now testing the adequacy of decision-making arrangements in traditional monopoly and restructured industries alike around the world.

Australia’s National Electricity Market (NEM) provides an interesting context for considering the challenges and options of facilitating wind energy integration in restructured electricity industries. The NEM has a large geographical scope, a rather different mix of generation from many other jurisdictions, and its own particular electricity market and associated renewable energy policy support framework. The last decade has also seen considerable wind energy development, some resulting integration challenges and significant consequent changes to NEM arrangements.

This paper aims to summarise some of the key design features of the NEM relevant to wind integration, its performance to date in facilitating appropriate wind deployment and possible changes to support much higher wind penetrations in the future. It first briefly summarises the current design of the NEM and its general performance to date. Section III assesses how well the current NEM arrangements might be expected to facilitate wind integration,

Impacts and best practices of large-scale wind power integration into electricity markets – some Australian perspectives

I. F. MacGill, Member, IEEE

W

978-1-4673-2729-9/12/$31.00 ©2012 IEEE

2

particular challenges that have emerged and possible changes to the NEM design. Section IV assesses how high wind penetrations in some regions of the NEM have already impacted on market outcomes both in terms of market prices, and also the revenues of market participants. It also briefly outlines some current investment challenges for wind. Section V then summarises some of the key lessons that the NEM holds for electricity market design to facilitate appropriate wind integration, and briefly flags some possible wider lessons for broader energy related policies including those to support greater renewable energy.

II. NEM DESIGN AND PERFORMANCE TO DATE The Australian NEM design is unusual amongst

restructured electricity industries around the world [1]. Its centre-piece is a set of compulsory regional gross-pool spot energy and Frequency Control Ancillary Services (FCAS) markets that solve a security-constrained dispatch every five minutes for five interconnected regions. All generators greater than 30MW in size (except for intermittent sources as will be discussed later) must participate in the NEM’s scheduling and cost allocation procedures. Both the system and market fall under the direction of the Australian Energy Market Operator (AEMO). Notably, the NEM has no formal (commercially accountable) day-ahead market or capacity market.

It has formal objectives of open access, ‘causer pays’ cost allocations and equal treatment, although these are difficult to deliver in practice within an electricity industry. Still, the general principle is that market participants pay, and/or are paid, according to their contribution to overall industry costs and benefits.

The wholesale spot market prices electricity every five minutes across five regions. FCAS markets determine frequency control ancillary services costs for regulation and contingencies, again at five minute intervals for the five regions. The NEM is therefore infused with uncertainty— generators can rebid within five minutes of dispatch and are highly motivated to respond to changing market conditions. A range of derivative markets that are not formally provided or supervised by NEM governance arrangements, offer opportunities to manage the price risk associated with these wholesale spot markets.

The NEM appears to have achieved reasonable success to date in matching commercial market signals with the underlying economics of the electricity industry, within an effective security regime. Wholesale prices have been generally low by international standards although a key factor here is the low costs of coal-fired generation in Australia, and the NEM reliability target has been met to date [2]. It has also successfully integrated wind generation approaching 2GW with over half of this wind located in the single, relatively small (peak demand of approximately 3.5GW) region of South Australia (SA) that now has a wind energy penetration of around 20%, and only limited interconnection with the rest of the NEM.

III. ELECTRICITY MARKET DESIGN TO FACILITATE WIND INTEGRATION

A. Key electricity market challenges with the large scale deployment of wind energy in the NEM

Challenges with high wind penetrations within electricity markets need to be seen within the context of wider electricity market objectives. Wind is proving one of the most assured and economically attractive options for reducing the environmental impacts and dependence on (often imported) fossil fuels in electricity industries around the world. By comparison, other low-carbon generation options such as Carbon Capture and Storage (CCS) and nuclear power have proven more difficult to deploy than many expected [3].

The challenges are, in the most general sense, the challenges of deploying a relatively novel generation source with rather different operational characteristics from existing options, into an industry providing an essential public good and posing complex operational challenges due to the nature of electricity itself, and the existing industry. It might be expected that similar challenges will emerge with other ‘disruptive’ technologies and this is, indeed, being seen with the growing deployment of solar technologies such as PV which raise some similar challenges.

Experience to date within the NEM and by comparison other electricity industries around the world would suggest that the NEM has some advantageous arrangements for effectively and efficiently facilitating the integration of significant levels of wind energy [1]:

Supply/demand balance for regulation, contingencies and energy is managed through a gross pool rather than primarily bilaterally. This may be advantageous for variable and somewhat unpredictable generation which can find it difficult to contract bilaterally forward on fixed volumes without finding itself having significant exposure to net balancing market outcomes. In the NEM, the spot market solves supply/demand balance for all generation and load every five minutes. .

The NEM provides transparent regional prices for all market participants that reflect a considerable aspect of the underlying locational, temporal and uncertainty value of electricity as it evolves over time. These price signals have significant implications for wind-farm investment as energy value represents a significant proportion of project revenue. In particular, the additional cashflow delivered to wind farms through the 20% renewable energy target does not shield wind farms from these signals in the way that feed-in tariff support arrangements may. Wind farms are therefore located with consideration of expected regional pool prices (that reflect potential inter- regional transmission constraints) and intraregional loss factors, and the predicted match between their generation and periods of generally higher pool prices. .

The NEM FCAS arrangements provide a highly transparent approach for pricing regulation and

3

contingency ancillary services which wind farms may have particular needs to call upon. There is considerable value in co-optimising energy and FCAS dispatch for intermittent generation.

The freedom of scheduled participants to rebid to the 5 minute dispatch boundary lets them revise offers with improving forecasting information down to near real-time and provides strong incentives for them to enhance their short-term operational flexibility. The use of shorter ‘gate closures’ also supports wind energy integration.

However, many challenges remain and new ones are emerging. Particular issues in the NEM include: - Providing the most useful forecasts of potential future wind

generation to assist all market participants. A key issue is forecasting possible extreme weather events that potentially threaten system security and therefore, given the NEM design, have potentially very significant commercial implications.

- The degree and nature of formal wind participation in market scheduling and cost allocations given its operational characteristics and capabilities. There are particular issues with the provision and payment of ancillary services, however, there are likely to be growing challenges in efficiently and securely managing supply-demand balance over the 5 minute to hours ahead time frame given potentially significant wind variability possible over such time frames and the limitations of conventional dispatchable plant in responding to such variability. .

- Transmission investment given that sites with good wind resources are often located in remote regions without strong network connection. Efficient use of existing network assets also would seem to require further attention in order to more efficiently manage congestion. .

B. Possible market design changes to facilitate wind

There are many emerging challenges and opportunities for electricity markets beyond wind generation including other intermittent and, furthermore, highly distributed renewables such as PV, the growing integration between gas and electricity markets and the need to facilitate far greater demand-side participation. These all need to be addressed in a coherent and comprehensive manner which, as far as possible, provides a technology and participant neutral basis for achieving both direct electricity industry objectives, but also wider societal objectives for the energy sector.

The last five years has seen a number of changes to NEM arrangements and rules to better facilitate higher wind penetrations including improved transparency, development of centralized Wind Energy Forecasting and changes to require wind generation to participate more formally in scheduling and security processes through a new semi-scheduled category of market participant. This latest development reflects the growing challenges of having significant intermittent generation within the NEM remaining outside some of the market’s important scheduling and security processes.

Particular areas receiving greater attention from the policy and rule making process at present include network investment and congestion and procurement of short-term reserves to meet security requirements [2]. More generally, wind energy will be facilitated through efforts to improve demand-side participation in the electricity industry where controllable loads can help manage ongoing supply-demand balance, as well as improvements to gas market arrangements that support greater use of highly flexible and hence complementary gas-fired generation.

More generally again, it is important that policy and rule makers revisit the governance arrangements in place for the electricity market design but also wider issues including industry structure and related markets such as those providing targeted policy support for renewables. Effective action on climate change will require ongoing changes at a speed and scale that may be beyond the present framework.

IV. HIGH WIND PENETRATION IMPACTS ON NEM OUTCOMES Whilst the NEM still has a relatively low wind energy

penetration of less than 3%, a number of regions have significant penetrations that provide some basis for assessing the impacts of high wind penetrations [4]. Considerable caution is, however, required given the many complex, uncertain and interacting drivers of market outcomes. Wholesale spot market prices and participant dispatch have very high transparency within the NEM, however derivative market prices are more important for many market participants in terms of driving investment and managing revenue risks associated with the spot market. Derivative prices and participant positions have far lower transparency than that seen for the wholesale market. There are also eight FCAS prices of potential relevance as well. All these prices vary by time and location, and the exposure of market participants to these varying prices depends upon their location and time changing generation.

Another key factor is that the NEM is an energy-only market and has no capacity markets as such. Instead, the market design permits very high prices in the short-term (two orders of magnitude greater than typical prices) at times of tight supply-demand balance. This provides a major incentive for generators to ensure they are available at such times. The risks associated with such price variability can be managed, at least in part, through derivative markets. The challenges wind energy poses in the capacity markets of some electricity industries might reflect the challenges of these capacity markets more generally - the electricity industry’s need for capacity depends on ongoing and potentially rapidly changing market conditions. A strength of the energy only market is that it rewards generation capacity available at the time it is required.

A. Large-scale wind integration impacts on NEM prices

Analysis of wholesale spot price outcomes over recent years in high wind penetration regions of the NEM highlights

4

that periods of high wind generation see typically lower prices than periods of low generation for all levels of demand (the most significant price driver) as shown in Fig. 1.

It is not clear what the overall impact of wind generation is on prices overall because of the many other drivers of price, and significant year on year variability that has been seen due to these. For example, SA annual volume weighted prices have varied from A$32/MWh to over A$100/MWh over the last decade. Reasons for this variation include extreme weather (particularly summer heat waves) as well as generation investment other than wind [2]. However, it is certainly apparent that periods of high wind generation in SA are associated with lower prices and, indeed, a growing number of negative price events [4].

These spot market outcomes highlight the importance of derivative market strategies for securing NEM participant returns. For a range of reasons, however, (notably the dominant market position of AGL in SA), derivative markets in SA exhibit relatively low liquidity by comparison with some other regions.

Fig. 1. Wholesale spot prices in the SA and VIC regions of the NEM as a function of demand for periods of top quartile and bottom quartile wind

generation at the time of that demand [5].

At present, turnover in the FCAS markets in the NEM is typically less than 0.5% of spot market turnover, reflecting at least in part the short time period over which these ancillary service markets operate given five minute spot market operation. ‘Causer pays’ principles for ancillary services in the NEM are challenging to implement in practice due to the shared nature of ongoing net supply

and demand imbalances within the industry. Current and potentially greater wind penetrations are not seen as likely to significantly increase contingency requirements, however, some work does suggest that high wind penetrations might drive significant increases in regulation service requirements and hence costs [6]. For wind farms, FCAS costs of currently A$0.40/MWh are low compared to typical spot market prices that are around one hundred times higher. However, causer pay requirements might see these increase significantly (perhaps ten fold) under higher penetrations

B. High wind penetration impacts on generation participants

Significant wind capacity investment will, as with any other generation technology investment, shift market outcomes and hence potentially impact on the revenue of all market participants. In a restructured electricity industry, such impacts are, of course, an intended outcome. However, the ‘external’ policy support for wind generation is certainly argued by some incumbent market participants to represent a market distortion.

The formal objective of the Australian NEM does not explicitly include any environmental, or for that matter social, goals. This was a design choice. Given that Australia’s energy policy objectives do include such objectives, external policies to change NEM outcomes in support of these objectives can actually be seen as an obligation rather than an imposition on the NEM [3].

As noted previously derivative markets play the key role in determining overall revenue for most market participants and the relationship between these prices and spot market prices is complex and uncertain. Nevertheless, it is interesting to note the spot price outcomes for wind and other generation seen in three NEM regions over recent years as wind penetrations have increased as shown in Figure 2.

Fig. 2. A comparison of volume weighted average annual prices received by

wind and non-wind generation in three regions of the NEM over 2009-10.

In SA and Victoria with significant wind penetrations, wind generation has on average been earning considerably less spot market revenue ($/MWh) than the other generation types. In SA certainly, this revenue difference has generally grown as wind penetrations have as well [4].

This has increasingly important implications for wind farm owners and operators, and potential investors within those regions. It reflects the reality that generation without inherent energy storage has lower value than conventional generation with storable primary energy sources such as coal, gas and

5

hydro. This earned price varies significantly between wind farms depending on how their particular wind regime matches underlying price patterns (typically higher prices in summer and winter afternoons and evenings) and the output of other wind farms (high wind output correlation with other wind generation means greater adverse merit order effects and hence typically lower prices at times of high output).

The participation of wind energy in the NEM’s associated derivative markets is made more challenging by the volume uncertainty associated with their operation. Wind farms might choose not to participate in these markets. However, there is also anecdotal evidence of volume firming derivatives being offered by some participants with flexible gas-fired plant for wind farms participating in the derivative markets. Although the volume risk poses significant challenges, there are good reasons for wind generation to participate and it has some potentially attractive capabilities by comparison with other generation.

For example, SA is seeing a growing number of negative price events [2]. Depending on their contracting arrangements, a number of windfarms are now reducing output when these negative prices fall below the renewable energy certificate price that they earn on their generation. Unlike coal-fired plant which can not reduce output below some minimum operating level (typically 40-50% of rated output), wind farms can reduce output rapidly and hence potentially profit significantly from such events when contracted.

C. Wind energy forecasting Since 2008 the NEM has had a centralized Australian Wind

Energy Forecasting System (AWEFS) to support security-driven and commercial decision making. AWEFS is designed to produce wind forecasts that can be integrated into AEMO’s forecasting processes, from the five minute dispatch process to the two year medium term projected assessment of system adequacy. It provides wind energy forecasts at individual windfarm, regional and system-wide aggregations including some measures of expected uncertainties

It is challenging to assess the value that the system provides in security and commercial terms. AEMO reports on forecast accuracy metrics, however, issues include both which measures of accuracy are the most relevant (for example, forecasting of extreme weather events may have far higher security and commercial value than forecast accuracy during times of normal market operation) and what value these forecasts actually provide. This is an important area for future work.

D. Wind generation investment The deployment of some 2GW of wind generation in the

NEM over the last decade highlights the success of NEM and wider policy arrangements to facilitate wind farm investment. The Federal Government’s 20% renewable energy target for 2020 is widely projected to drive considerable wind generation investment over the coming few years [2]. However, there are some current challenges and uncertainties

that are adversely impacting wind farm development. In large part these would seem to result from the poor design of the renewable energy target which has seen unexpectedly high deployment of small-scale PV systems and even solar hot water systems over the past few years provide ‘deemed’ renewable energy generation that exceeds the scheme targets for the next several years.

However, there are growing concerns regarding transmission arrangements, particularly given the time lines required for major network development. Likely most importantly, there are also emerging structural challenges within the NEM that are influencing investment. Through both horizontal and vertical integration over the past decade, three large and vertically integrated utilities now own and operate 30% of NEM generation and supply over 80% of small retail customers. They are also responsible around 60% of the generation investment seen since 2007. These three firms are now the key players in whether and how the renewable target might be achieved.

V. CONCLUSIONS The NEM would appear to have reasonably effectively and

efficiently managed regionally significant, although NEM- wide still relatively modest, wind penetrations. Wind generation would now seem to be having some significant impacts on market outcomes in several market regions. In particular, wind generation in regions with high wind penetrations is receiving reduced market revenues (A$/MWh) than other generation. This is providing an increasing incentive for wind-farm developers to look for project opportunities in other States that may not feature as attractive wind resource or site availability, but which have a lower wind penetration. This is, of course, a better industry outcome in terms of managing the integration costs and other challenges associated with high penetrations of renewable energy sources that do not have inherent energy storage.

Despite these achievements, projected significant increases in wind penetration due to federal government renewable energy targets for 2020 will continue to test the adequacy of market arrangements and likely require ongoing market design changes. These changes need to be seen in the wider context of NEM challenges in transitioning towards a low-carbon and hence more sustainable electricity industry future.

A larger market and wider policy design challenge would actually appear to be the low-carbon energy technology policy support that will be required to deploy disruptive low carbon technologies. Designers of the market-oriented policy support measures such as the renewable energy target scheme in Australia might well learn from electricity market design on the importance of - high transparency and disclosure by market participants - robustness against unexpected developments and market

participant behaviours (including during the design process) as seen with the close integration of commercial and security arrangements within the NEM, and

- governance arrangements including formal separation of

6

powers and interfaces between policy making rule making, operation and enforcement, as well as formal rules for changing the rules in response to changing market outcomes as seen with Australian NEM rule change process.

VI. ACKNOWLEDGMENT The author gratefully acknowledges the many and varied

contributions of colleagues to the work that has been undertaken at the University of NSW Centre for Energy nad Environmental Markets (CEEM) on the challenges and opportunities of renewable energy integration. Particular thanks are owed to Professor Hugh Outhred and Dr Nicholas Cutler. Numerous research students have also contributed to this work including Nicholas Boerema, Sam Forrest and Sebastian Oliva. This work is supported in part by Australian Research Council and Australian Solar Institute funding.

VII. REFERENCES [1] I.F. MacGill, “Electricity market design for facilitating the integration of

wind energy: Experience and prospects with the Australian National Electricity Market,” Energy Policy 38 (2010) p. 3180–3191.

[2] Australian Energy Regulator (AER), “State of the Energy Market 2011,” Available at www.aer.gov.au.

[3] International Energy Agency (IEA), “World Energy Outlook” (2011), available at www.iea.org.

[4] N. Cutler, N. Boerema, I.F. MacGill and H.R. Outhred, “High penetration wind generation impacts on spot prices in the Australian national electricity market, Energy Policy 39 (2011) p. 5939-5949.

[5] S. Forrest, “Quantifying the Impact of Intermittent Wind Generation in [6] the Australian National Electricity Market,” Engineering Honours

Thesis, University of NSW, June 2011. [7] Roam Consulting, “Impact of the LRET on the costs of FCAS, NCAS

and Transmission augmentation,” Report to the Australian Energy Market Commission, September 2011, available at www.aemc.gov.au.

VIII. BIOGRAPHIES

Iain MacGill is an Associate Professor in the School of Electrical Engineering and Telecommunications at the University of New South Wales, Sydney, Australia, and Joint Director for the University’s interdisciplinary Centre for Energy and Environmental Markets. His teaching and research interests include electricity industry restructuring and sustainable energy technologies, with a particular focus on distributed resources and energy policy.

862 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

Integration of Wind Generation Into the ERCOT Market

Henry L. Durrwachter, Senior Member, IEEE, and Sherry K. Looney

Abstract—This paper examines the market rules implemented within the Electric Reliability Council of Texas (ERCOT) zonal and nodal market designs related to the integration of wind gen- eration. Currently, there is almost 10000 MW of wind generation installed and operating in the ERCOT market. Total ERCOT summer peak load is in excess of 65000 MW and wind generation penetration has exceeded 25% on several instances. This paper discusses the various rules and market mechanisms put in place to effectively deal with a high penetration of variable wind gener- ation while maintaining adequate system reliability and creating proper market price signals for wind generation. Specifically, the paper includes treatment of wind generation in both the zonal and nodal markets, including day-ahead markets and real-time operations and describes the actual market outcomes which have been realized through several years of increasing penetration of wind generation on the ERCOT system.

IndexTerms—Ancillaryservicemarkets,bilateralmarkets,elec- tricity markets, market design, market operation, variable gener- ation, wind generation.

I. INTRODUCTION

A COMPETITIVE retail electricity market became realityin the Electric Reliability Council of Texas (ERCOT) on January 1, 2002 [1]. Since the initial implementation of that zonal market, significant amounts of variable wind generation resourceshaveaddedtotheportfolioof resources in theERCOT system as shown in Fig. 1. The impetus for theadditionof thesenewwindgeneration re-

sourceswasprovidedbyaRenewablePortfolioStandard (RPS) implemented by the Texas Legislature in 1999 that initially re- quired that2000megawatts (MW)ofnewrenewablegeneration capacity be added in Texas by January 1, 2009 [2]. That RPS was subsequently changed to 5000 MW by January 1, 2015. Today, there is almost 10000 MW of wind generation installed and operating in the ERCOT market. The integration of increasing quantities of wind generation

resources requires increasingly complex analysis and impacts the design of many operational systems. Some of the most sig- nificant concerns associatedwith the integration of wind gener- ation in the ERCOT system have been transmission adequacy and congestion, reactive and primary frequency response, and adequacy of ancillary services plans.

Manuscript received September 16, 2011; revised May 11, 2012; accepted May 28, 2012. Date of publication July 25, 2012; date of current version September 14, 2012. The authors are with the Department of Regulatory, Luminant Energy Com-

pany LLC, Dallas, TX 75201 USA (e-mail: [email protected]). Color versions of oneormore of thefigures in this paper are available online

at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TSTE.2012.2202697

Fig. 1. Installed wind generation capacity in ERCOT.

ERCOTand thePublicUtilityCommissionofTexas (PUCT) completed an exhaustive transmission study to develop the Competitive Renewable Energy Zones (CREZs) [3]. The CREZs were created to determine the likely location of addi- tional wind generation and to provide a means of determining the amount of additional transmission needed to deliver wind power to ERCOT loads. However, the additional transmission capacity resulting from the CREZ construction will not be completed until the 2013 to 2014 time frame. The effects of inadequate transmission capacity have been

addressed in the short term by changes in market system rules to provide equitable dispatch, i.e., curtailment, of wind generation resources to control local and interzonal congestion whilemaintainingminimumreliability standardsestablishedby ERCOT operating as the Independent System Operator (ISO). As wind generation capacity increased during the ERCOT zonal market operation, rules and operational changes were required to manage the increasing amount of wind curtailments needed to manage congestion and maintain system reliability. Provisions for reactivepower support andprimary frequency

response from wind generation resources have been addressed through ERCOT market participant and regulatory processes. The reactive requirements for wind generation resources were initially proposed through zonal Protocol Revision Request (PRR) 830, which was approved by the ERCOT Board and subsequently appealed to the PUCT by a group of wind gen- erator owners. The proposal sets a minimum standard for reactive voltage support from wind generators which can be accomplished through dynamic ratings and static VAR devices. Upon final dispositionof the appeal by the PUCT, the approved requirements will also be incorporated into the ERCOT nodal

1949-3029/$31.00 © 2012 IEEE

DURRWACHTER AND LOONEY: INTEGRATION OF WIND GENERATION INTO THE ERCOT MARKET 863

Fig. 2. Monthly energy from wind as a percentage of total ERCOT energy production.

protocols via a Nodal Protocol Revision Request (NPRR). Pro- tocol revisions also added a requirement for wind generators which connected after January 1, 2010 to have an adjustable dead band and provide primary frequency response similar to the droop characteristic of five percent used by conventional electric generators. On December 1, 2010, the ERCOT market changed from a

zonal market to a nodal market design [4]. This paper provides a brief review of the current status of the wind generation in ERCOT. This paper also examines the market rules put into place, first in the ERCOT zonal market and later in the ERCOT nodalmarket, toallowtheeffective integrationof largeamounts of variable wind generation resources in both day-ahead and real-timemarkets.Andfinally, this paper examines themethod- ology fordeterminingancillary service requirements associated with increasing quantities of wind generation.

II. STATUS OF WIND GENERATION CAPACITY IN ERCOT

At the end of June, 2011, ERCOT had approximately 9452 MW of wind generation capacity [5]. The monthly amounts of wind energy production as a percentage of total monthly energy production in ERCOT for the years 2010 and 2011 are shown in Fig. 2. Furthermore, the total amount of energy in megawatt hours

(MWh) produced annually by wind generation in ERCOT con- tinues to increase as shown in Fig. 3.

III. ZONAL MARKET OPERATION OF WIND GENERATION RESOURCES

Wind generation resources (WGRs) were initially incorpo- rated into the ERCOT system through a zonal market struc- ture. In the beginning, when the quantities of wind generation were low,WGRswere treatedasnoncontrollablegenerationand price-takers in themarket.As thewindgenerationquantities in- creased in theWest congestionzone,market rules changeswere approvedby themarketparticipants toaddressoperatingandre- liability concerns. At first, areas of local transmission congestion were ad-

dressed with preassigned, prorated capacity rights to equitably manage transmission capacity among competing WGRs. As additionalWGRcapacitywasadded to thesystem, transmission congestion occurred at commercially significant constraints

Fig. 3. Cumulative energy from wind in ERCOT.

(CSCs) on the 345-kV interconnections between the ERCOT West congestion zone and each of the other market congestions zones, (i.e., North, South, and Houston). In the ERCOT zonal market, balancing energy was deployed

from an up and down bid energy stack during each 15-min set- tlement interval in order to manage the projected generation load for the next 15 min. When energy flows on the CSCs be- tween zones were below the operating limits, the balancing en- ergywasdeployedsystem-wideandall zonalpriceswereequal. However, when any one of the CSCs became limiting, the bal- ancing energy was deployed among zones and the zonal prices diverged, i.e., prices in “generation pockets” went down and prices in “load pockets” went up. As ERCOT began encoun- tering significant amounts of congestion from the West Zone (causedbythe large influxofwindgeneration in thatzone),gen- eration had to be decreased to manage the CSC limits. Initially, WGRs were allowed to produce at full capability

with wind variability. The ERCOT Zonal Market employed a 15-min interval for balancing generation to load. The sched- uling, pricing, and dispatch (SPD) software was executed for minimumcostdispatchevery15minfor thenext interval’s fore- casted load. Each qualified scheduling entity (QSE) submitted up balancing energy services (UBES) and down balancing en- ergy services (DBES) offers for those generation units in the QSE’sportfolio.TheUBESandDBESofferswereonaconsol- idatedportfolio basis so thatwhen theQSEreceivedaUBESor DBES deployment for each congestion zone for each interval, the QSE could optimally generate from any unit to fulfill the deployment. ERCOT had traditionally used a zonal minimum down-balancing requirement applied to the conventional gen- eration portfolio of each qualified scheduling entity (QSE) as a method for ensuring the ability to manage decreasing load sce- narios, i.e., when generation on the system should decrease for powerbalance.TheWGRswereinitiallyexemptfromthisdown balancing requirement. As the need to curtail WGRs increased, ERCOT initially

managed this need with out of merit energy (OOME) instruc- tions to the individual WGRs. This required numerous manual interventions by the ERCOT operators. Due to widespread curtailment of wind generation during 2008, West Zone prices were frequently set at negative values due to the quantity of DBES required to manage the congestion. The negative prices were a reflection of the effect of the federal Production Tax Credit (PTC) which resulted in a negative break-even price

864 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

Fig. 4. Number of congested intervals per quarter.

Fig. 5. West Zone real-time price distribution during congested hours.

to generators. In January, 2009, ERCOT market rules were revised to require WGRs to set their low sustainable limit (LSL) to 10% of their installed capacity and begin offering the minimum down-balancing requirement based on the difference between the high sustainable limit (HSL) and LSL [6]. This re- sulted in a more operationally systematic solution to interzonal congestion. Fig. 4 shows thenumberof settlement intervalsperquarter in

whichtheWestZonewindproductionwascurtailedduetozonal congestion resulting in negative West Zone real-time Market Clearing Price of Energy (MCPE) for the years 2007 through 2010. With few exceptions, the curtailed intervals resulted in negativeWestZoneprices, as shownonFig. 5.Fig. 6 shows the MCPEs which resulted from operating the zonal market during theperiodof2007 through2010whenwindgenerationcapacity increasedsignificantly.Thenumberofcongested intervalsactu- ally decreased from 2008 to 2009 due to transmission upgrades which increased the transfer limits between the West and North congestion zones. However, the large number of intervals with negative pricing and significant price separation between the West Zone and all othermarket zones persisted through the end of the zonal market operation on November 30, 2010.

IV. INTEGRATION OF WGRs INTO THE NODAL MARKET STRUCTURE

Upon implementation of the ERCOT Nodal market on De- cember 1, 2010, transmission congestion was managed by dis- patching each WGR based on its Energy Offer Curve. All gen- eration resourceswithinERCOT,whenonline andavailable for energydispatch, are required to have either anOutputSchedule or anEnergyOfferCurve for the specificunit.Thesecuritycon- strained economic dispatch (SCED) algorithm in the ERCOT

Fig. 6. Average annual zonal real time energy price.

Nodal market optimally dispatches each unit’s energy in inter- valsof5minor less tomatchaggregate systemloadwhileman- aging congestion limits. The SCED algorithm relies on the net- work topologyprovidedbyadetailednetworkmodel that accu- ratelydepicts thenear real-time transmission capacity available throughout the system, allowingERCOTtodispatch theWGRs more efficiently. However, the negative pricing for wind gener- ation during certain periods will continue to persist until addi- tional transmission capacity is realized through the completion of the CREZ projects. In the initialdevelopmentof theNodalmarket rules, itwasas-

sumedthat thehighsustainable limit (HSL)of theWGRswould be set to their actual output level based on telemetry so that SCEDwouldhave thebest estimateof the actual production for the next SCED execution. However, as the need for WGR cur- tailment became apparent in the zonal market, setting the HSL equal to the actual energy production from the wind generation resource was not indicative of the actual production potential due to the reduction in actual energy production as a result of curtailment. Thus, the HSL was limited due to the curtailment. When not curtailed, the WGR is required to continue to

telemeter the actual production as its HSL. Initially, when the dispatch level (referred to as the Base Point in the Nodal market protocols) received from SCED was less than the actual wind generation production by more than 2 MW, the WGR was considered to be curtailed because SCED had calculated a lower base point than full output due to congestion limitations on the dispatch. However, determining systematically exactly when the WGR was curtailed was problematic because too many variables were changing in real time. Therefore, ERCOT made a change in their system to telemeter a curtailment in- dicator or “flag” to the control system of the QSE responsible for responding to the ERCOT dispatch instruction to the WGR. When the WGR is curtailed by SCED to manage congestion, the HSL telemetered from the WGR is set to a calculated wind generation resource production potential (WGRPP). The WGRPP is calculated by the WGR based on wind speed at the plant site using empirical data correlation. When the curtailment ends, the WGR is required to return to telemetering the HSL equal to the actual wind generation (after a delay to allow for the effects of the curtailment to end) so that HSL

DURRWACHTER AND LOONEY: INTEGRATION OF WIND GENERATION INTO THE ERCOT MARKET 865

Fig. 7. Actual wind production and estimates of curtailment.

telemetered value represents the actual physical uncurtailed WGR generation. Fig. 7 shows actual wind production and estimates of local,

zonal curtailment for 2008 through November 2010. The data shown for 2011 shows only wind production. Data on curtail- ments under the nodal market design in 2011 are not currently available.

V. METHODOLOGY FOR DETERMINING ANCILLARY SERVICE REQUIREMENTS WITH INCREASING QUANTITIES

OF WIND GENERATION

ERCOT contracted with GE Energy to produce an Analysis of Wind Generation Impact on ERCOT Ancillary Services Re- quirements in 2008, referred to hereafter as the 2008 GE Wind Study [7]. Prior to 2008, ERCOT had used a methodology for determining the quantity of up regulation service (URS) and down regulation service (DRS) by blocks of hours for each day of the month. The methodology required ERCOT to calculate the average and standard deviation for the URS and DRS deployed for each hour in the previous month and for the same month of the previous year. For each of these months, ERCOT then calculated the amount of regulation service required by hour to provide an adequate supply of regulation capability 98.8% of the time, or effectively 2.5 standard deviations [8]. A key component for successful integration of large quanti-

ties of wind generation into ERCOT was and continues to be

provision of adequate quantities of ancillary services to com- pensate for large, unpredictable wind ramping events. Every year, ERCOT has made incremental improvements to its an- cillary service requirements methodologies to address the chal- lenges of increasing wind penetration in the market. In 2008, theblockhour concept of regulation requirementswas replaced by individual hourly requirements. ERCOT also started per- forming a back-cast of the regulation exhaustion rate to deter- mine how much additional regulation would be needed in the previous month to avoid exceeding a 1.2% exhaustion rate for any hour. In 2009, ERCOT began using tables from the 2008 GEWindStudy toestimate the increase in regulationneededfor increasing wind generation capacity during the year. In 2010, ERCOT added the provision for an additional regulation incre- mental requirement if the average 1-min Control Performance Standard (CPS1) score was less than 90 or 100 for the previous month. All of these successive changes were made to continue to optimize the regulation requirements on the system. Forpurposesofcalculationofancillary service requirements,

ERCOT treats wind generation as “negative load.” Ancillary service responsibilities resulting from uncertainty in load fore- castingalsoconsiderseparately theuncertaintyof thewindfore- cast itself.Conventionalgenerationcapacity(i.e.,generatingca- pacity other than wind generation) is dispatched to meet the net load which is defined as firm customer load less actual wind generation. In2011,ERCOTfurther increased thegranularityof

866 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 4, OCTOBER 2012

Fig. 8. Hourly URS requirements for the month of April.

Fig. 9. Hourly DRS requirements for the month of April.

the calculationof regulation requirements by adding criteria re- quirement that theregulationforeachhourmustbeat leastequal to the greatest 5-min net load change for the same hour of any day of the previous month. This progression of improvements to the regulation requirement calculations shows the proactive response ERCOT has made to optimize the procurement of an- cillary services as thewindgenerationhas steadily increasedon the system. Wind generation in West Texas is anti-correlated or out-of-

phase with the daily load curve (i.e., the maximum output of windgeneration inWestTexasgenerally occurswhencustomer loads are the lowest). In ERCOT’s experience, load and wind forecasterrorsarevirtually independentandnet loadforecasting accuracy decreases with increasing wind penetration. One would expect that the requirement for regulation ser-

vice on the ERCOT system would increase significantly with increasing wind penetration. However, with the optimization of the requirement methodology, the increase in quantities of regulation purchased has actually been minimal. For the month of April, the increase in wind generation was 79% from 2007 to 2008, 40% from 2008 to 2009, and 28% from 2009 to 2010. Figs. 8 and 9 show the hourly URS and DRS requirements for the month of April, 2007 through 2010. The URS and DRS requirements increased in some hours and decreased in other hours year to year, but overall, did not significantly change even though the amount of installed wind generating capacity in ERCOT significantly increased over the same time period.

Fig. 10. Wind output, regulation, and RRS deployments, January 28, 2010.

Based on meteorological analysis, the 2008 GE Study pre- dicted there could be a 2000- to 3000-MW decrease in wind generationovera30-minperiodunder thehighwindgeneration capacity scenario. This type of large change in wind generation requiresadequateresponseofallancillaryservice typestomain- tain reliability standards and grid stability. Fig. 10 shows the actual wind output, regulation and responsive reserve service (RRS), i.e., spinningreserve,deployments forJanuary28,2010. On that day, ERCOT experienced an actual decrease in wind generation of 2500 MW in approximately one hour. As shown on Fig. 10, at 0930 and 1400 on January 28, 2010, URS was fully deployed and RRS was deployed to maintain frequency. Since Nonspin Reserve Service (NSRS) in ERCOT is pro-

cured for load uncertainty and wind is considered to be “nega- tive load,” large changes in wind generation (often referred to as “ramping events”) must be considered in the NSRS require- mentcriteria.ERCOTincorporated the recommendationsof the 2008GEStudy into itsmethodologyfordeterminationofNSRS requirements byfirst calculating anet loadvalue by subtracting the actual wind generation from the actual load. The historical net load is then compared to the forecasts of customer load and wind generation output to determine the historical accuracy ob- served in forecasting. ERCOT then subtracts the up regulation service requirement from the 95th percentile of the net load un- certainty to determine the amount of NSRS to purchase during each hour of the day for the upcoming month. For on-peak hours, i.e., hours ending 7 through 22, ERCOT sets a floor on the NSRS requirement equal to the capacity of the largest con- ventional generating unit. By appropriately distributing the ad- ditional reservescoincidentwith the risk factorsassociatedwith specific timeperiods,bothseasonallyandona timeofdaybasis, the quantities of ancillary service requirements are optimized. Improvements in wind generation forecasting will help

optimize system control performance and ancillary service procurements.ERCOTcurrently utilizes a state-of-the-artwind forecasting model that applies artificial intelligence techniques to “learn” from its mistakes to continually improve both its hour-ahead and day-ahead wind production forecasts. The mean absolute percent error in the day-ahead and hour-ahead forecasts is currently approximately 10% and 7%–8%, respec- tively. ERCOT has found it difficult to correlate the actual generation of WGRs with the forecast of WGR production due to the significant durations of curtailments that are continuing

DURRWACHTER AND LOONEY: INTEGRATION OF WIND GENERATION INTO THE ERCOT MARKET 867

to be experienced. For example, during periods when the WGR limits production due to curtailments, the correlation of actual wind togenerationcannotbe included in theempirical analysis. In addition, ERCOT has recently implemented a tool de-

veloped in conjunction with its wind forecasting contractor to alert the ERCOT system operators to predict impending large ramping events [9]. Improvement in short-term wind forecasting will increase the effectiveness of this tool. With the recent implementation of a curtailment indicator flag so that curtailment periods are more accurately accounted for, wind generators will now have information necessary to allow them to more accurately analyze operational data and improve their wind generation forecasts.

VI. CONCLUSION

Since 2001, ERCOT has successfully integrated almost 10000MWofwindgeneration resources into a relatively small system (68000 MW peak load) through continuous modifica- tion and improvement of its market rules [10]. That success has been achieved by continual review and revision of market rules by stakeholders and regulators alike. New tools and operational procedures have been developed and implemented to allow ERCOT to more efficiently and more reliably manage an ever-increasing penetration of variable wind generation in its system. However, as more wind generation is added to the ERCOT system, additional challenges lie ahead.

REFERENCES

[1] Public Utility Regulatory Act Title II, Public Utilities Code, Chapter 39, Restructuring of Electric Utility Industry, Subchapter A—General Provisions, Section 39.001(b).

[2] Public Utility Regulatory Act Title II, Public Utilities Code, Chapter 39, Restructuring of Electric Utility Industry, Subchapter Z—Miscel- laneous Provisions, Section 39.904(a).

[3] Competitive Renewable Energy Zones (CREZ) Transmission Op- timization Study Public Utility Commission of Texas Docket No. 33672, Apr. 2, 2008, .

[4] ERCOT Market Notice General, Remaining effective dates of certain Nodal Protocols and Operating Guides Sections and retirement dates of Certain Zonal Protocols and Zonal Operating Guides in preparation for fullNodalMarketOperationsonOperatingDayDecember1,2010, October 20, 2010.

[5] ERCOT System Planning Report , Jul. 2011. [6] ERCOT Zonal Protocols Section 4.10.4, Resource Low Sustainable

Limit as aPercent of High Sustainable Limit Measure [Online].Avail- able: www.ercot.com

[7] Public Utility Commission of Texas Docket No. 33672 Mar. 28, 2008, Analysis of Wind Generation Impact on ERCOT Ancillary Services Requirements, GE Energy.

[8] 2010–2011ERCOTMethodologies forDeterminingAncillaryService Requirements ERCOT Board Approved 11/16/2010.

[9] ERCOT Press Release Mar. 25, 2010. [10] ERCOT Press Release Aug. 3, 2011.

Henry L. Durrwachter (M’73–SM’89) received the B.S.E.E. degree in electrical engineering from the University of Texas at Arlington in 1972. He is currently Director of ERCOT Market

Services at Luminant Energy Company LLC. He has served on the Board of Directors of the Texas Renewable Energy Industries Association (TREIA), the Dallas Electric Club, and currently serves as President of the Board of Directors of the Utility Wind Integration Group (UWIG). He is a Registered Professional Engineer in the State of Texas and has

over 39 years of experience in the electric power industry. Mr.Durrwachter is amemberof the IEEEPowerEngineeringSociety (PES).

SherryK.Looney received theB.S.C.H.E.degree in chemical engineering from the University of Texas, Austin, in 1980. She is currently a Sr. Project Manager in Regu-

latory Affairs at Luminant Energy Company LLC. She also serves as the Chair of the Qualified Sched- ulingEntitiesManagersWorkingGroupaspartof the ERCOTmarketparticipantorganization.Sheandhas 30years experience inoil andgasproduction,project engineering, gas pipeline operation, and power mar- keting operations.

Risk Modeling in Strategic Behavior and Equilibrium of Electricity Markets with Wind Power

Kang Xiaoning, Wang Xian, Zhang Shaohua, Key Laboratory of Power Station Automation Technology, Department of Automation,

Shanghai University, Shanghai 200072, China E-mail: [email protected]

Abstract—High wind power penetration in power systems will significantly increase risks faced by the conventional generators in the deregulated electricity markets. This will further affect these generators’ risk preferences and strategic behaviors. A supply function equilibrium model of electricity markets with wind power is developed taking into account the conventional strategic generators’ risk preferences. The impacts of generators’ risk preferences on the strategic behaviors and market equilibrium are theoretically examined in detail. Numerical examples are presented to verify the validity of the theoretical analysis. It is shown that through strategic bidding, a generator’s output and profit risk will be lowered with increasing its risk aversion. In addition, a generator’s expected profit and its risk will increase with increasing its rivals’ risk aversion.

Keywords-electricity market; wind power; risk preference; supply function equilibrium

I. INTRODUCTION The worldwide restructuring and deregulation of electric

power industries has been accompanied by extensive research on strategic behaviors and market power analysis of oligopolistic electricity markets. Equilibrium models using game-theoretic behavioral assumptions are broadly employed to examine strategic interactions among participants in electricity markets [1], [2]. In recent years, wind power, due to its significant role in mitigating environment pollution, has been widely used and rapidly developed around the world [3], [4]. Wind power has features of relatively strong randomness, volatility and intermittence [5]. Thus large-scale wind power penetration in power systems will inevitably increase risks faced by conventional generators who participate in market competition. This will further affect these generators’ risk preferences and strategic behaviors in electricity markets. In this context, integration of risks in equilibrium models of electricity markets with wind power is identified as an important issue.

Until now, on the one hand, a considerable amount of research has been conducted on risk modeling in single generator or purchaser’s optimization problem in the deregulated power market. For example, in order to model the generator or purchaser’s risk preference, the mean-variance utility theory is used in [6]-[12] while in [13]-[15] the conditional value at risk (CVaR) theory is employed. On the other hand, much research work has also been conducted regarding equilibrium models of electricity markets. Among

the most extensively used models are the Cournot model [16]-[20] and the linear supply function equilibrium (LSFE) model [21]-[26]. It can be noted that to date, in most of the equilibrium models of electricity markets, wind power is not taken into account. In addition, the conventional generators’ risk preferences are overlooked, or all generators are assumed to be risk neutral in the presence of risks.

Given this background, a LSFE model of electricity markets with wind power is presented in this paper. The mean-variance utility theory is employed to model the conventional generators’ risk preferences. The impacts of generators’ risk preferences on the strategic behaviors and market equilibrium are theoretically examined in detail. Numerical examples are presented to verify the validity of the theoretical analysis.

II. THEORETICAL MODEL

A. Assumptions Suppose that in a power market, there are n strategic

conventional generators and a certain number of wind power units. The wind power units are assumed to be price-takers. The market demand at time period t is expressed by the following linear inverse demand function:

p a bD= − (1) where, p is the market price at time t; a and b are constant coefficients taking values greater than 0; D is the market demand at time t.

The conventional generators have the following quadratic cost functions:

2( ) 0.5i i i i i iC Q Q Qα β= + ，i=1,2,…,n (2) where，Qi is generator i’s output; αi and βi are cost parameters taking values greater than 0.

The output of wind power units in time t, Qw, is assumed to follow a probability distribution with a mean value of μw and a standard deviation of σw. Since the market demand D satisfies:

1

n

i w i

D Q Q =

= +∑ (3) the inverse demand function in (1) can be reformulated as follows:

This project is supported by National Natural Science Foundation of China (No.70871074) and "11th Five-Year Plan" 211 Construction Project of Shanghai University.

978-1-4244-6255-1/11/$26.00 ©2011 IEEE

1 ( )

n

i i

p a b Q Qω =

= − +∑ (4)

The conventional generators compete by submitting their bids in the form of linear supply functions as follows:

i i iQ x d p= + ，i=1,2,…,n (5) where, xi and di are the intercept and slope of supply function respectively; p is the expected market price. In this paper, the parameter xi is chosen as generator i’s strategic variable, and di is assumed to be the same as the slope of generator i’s marginal cost, i.e. di=1/βi.

B. Equilibrium Model The profit of generator i is equal to the payment for its

production quantities at the market prices, minus its generation costs, that is:

( )i i i ip Q C Qπ = ⋅ − (6) The mean and variance of generator i’s profit can be

calculated as follows: 2E[ ] 0.5i i i i i ip Q Q Qπ α β= ⋅ − − (7)

2 2 2 2Var[ ] Var[ ]i i i wQ p Q bπ σ= ⋅ = (8) Using the mean-variance utility theory, generator i’s

decision problem can be formulated as the following utility-maximization problem:

2 2 2 2

M ax (1 )E[ ] Var[ ]

(1 ) ( ) 0.5 i

i i i i ix

i i i i i i i

J r r

r p Q Q r b Q ω

π π

α β σ

= − −

⎡ ⎤= − − − −⎣ ⎦ (9)

subject to

1 ( )

n

j j

p a b Q ωμ =

= − +∑ (10) , 1, 2, ,j j jQ x d p j n= + = (11)

where，Ji is generator i’s utility function；ri is generator i’s risk preference factor. ri=0 means that generator i is risk neutral, 0<ri<1 indicates that generator i is risk averse and the degree of generator i’s risk aversion will increase with increasing ri.

The equilibrium model taking into account generators’ risk preferences can be formulated by gathering n generators’ decision problems expressed by (9)-(11). By combining each generator’s first-order optimality (KKT) conditions, the equilibrium solution of the model can be obtained by solving these conditions.

From (9)-(11), the KKT conditions of generator i’s decision problem can be derived as follows:

2 2

1

(1 )[ ( ) ]

2 ( 1) 0

i i i i i i i n

i i i j j

r Q d p Q d

r b d Q b dω

α β

σ λ =

− + − −

− − + =∑ (12)

2 2(1 )[( ) ] 2 0i i i i i ir p Q r b Q bωα β σ λ− − − − − = (13)

1 0

n

j j

a b Q pωμ =

⎡ ⎤ − + − =⎢ ⎥

⎣ ⎦ ∑ (14)

, 1, 2, ,j j jQ x d p j n= + = (15) where, λ is the Lagrange multiplier associated with (10).

From (12) and (13), the equilibrium output of generator i(i=1,2,…,n) can be obtained by eliminating λ:

(1 )( )( )i i i i

i

r p bd G Q

F α− − −

= (16)

where, 2 22 ( ) (1 )i i i i iF rb bd G G rωσ β= − − − (17)

1

1 n

j j

G b d =

= +∑ (18) Substituting (16) into (14), the equilibrium of expected

market price can be derived as follows:

1

1

1

n j j

j j n

j

j j

H a b b

F p

H b

F

ω

α μ

=

=

− + =

+

∑

∑ (19)

Where, (1 )( )j j jH r bd G= − − ，j=1,2,…,n (20) Generator i’s equilibrium output Qi (i=1,2,…,n) can be

obtained by applying (19) to (16). Using (15), generator i’s strategic variable xi can be easily calculated as:

i i ix Q d p= − ，i=1,2,…,n (21)

C. Impacts of Generator’s Risk Preference on Equilibrium In this section, the impacts of generators’ risk preferences

on the equilibrium results, such as generators’ strategic variables, expected profits and the expected market price, are examined by theoretical analysis.

From (17)-(20), the first-order partial derivative of the expected market price with respect to generator i’s risk preference can be derived as follows：

3 2 2

2

1

2 ( ) ( )

(1 )

i in

ji i

j j

b G bdp p

Hr b F

F

ωσ α

=

−∂ = ⋅ −

∂ + ⋅∑

(22)

From (16)-(20), the first-order partial derivative of generator i’s equilibrium output with respect to its risk preference can be expressed as follows.

2 2 2

1,2

1

2 ( ) (1 )( )

(1 )

n ji i

in j j j ii j

i j j

HQ b G bd b p

Hr F b F

F

ωσ α = ≠

=

∂ − = ⋅ + −

∂ +

∑ ∑

(23)

Using (21), the first-order partial derivative of generator i’s strategic variable with respect to its risk preference is:

i i i

i i i

x Q p d

r r r ∂ ∂ ∂

= − ∂ ∂ ∂

(24)

From (18), it can be found that 1

1 0 n

j i j

G b d bd =

= + > >∑ , this implies that 0ibd G− < . Given that 0 1ir≤ < , it can be observed from (17) and (21) that Fj<0 and Hj<0, j=1,2,…,n. Due to the fact that in general, Qi≥0, thus it can be seen from (16) that ip α≥ . Therefore, from (22) and (23), it can be concluded that 0ip r∂ ∂ ≥ and 0i iQ r∂ ∂ ≤ . As such, it can be followed from (24) that 0i ix r∂ ∂ ≤ .

The above theoretical results show that generators’ risk preferences have important influences on their strategic behaviors and the market equilibrium. A generator’s output will decrease with increasing its risk aversion and this is achieved by reducing its strategic variable. As a result, the expected market price will increase with increasing generators’ risk aversion.

In the following, the impacts of each generator’s risk preference on its expected profit and variance are further examined. From (7), the first-order partial derivative of generator i’s expected profit with respect to its risk preference can be derived as follows.

E[ ] [ ( )]i ii i i i

i i i

Qp Q p Q

r r r π

α β ∂ ∂∂

= ⋅ + − + ∂ ∂ ∂

(25)

From (12) and (13), the following equation can be obtained by eliminating λ:

2 2

1,

2 ( )

1 1

i i i i i i n

i j

j j i

r b Q Q b p Q

r b d

ωσα β

= ≠

− + = + − +∑

(26)

From (22) and (23), we have:

1,

1 n

j

j j i ji

i i

H b

FQ p r r b

= ≠

+ ∂ ∂

= − ⋅ ∂ ∂

∑ (27)

Substituting (26) and (27) into (25) gives: E[ ]

(1 )i i i i

p Q K

r r π∂ ∂

= ⋅ − ∂ ∂

(28)

where,

2

1,

1

2 1

1,

1,

2 1 1 1

12 1

1 (1 )

i n

i j

j j i

n

kn j k

n j j i j

k j k k j

r b K

r b d

b dr b r b d d b

ω

ω

σ

σ

= ≠

−

=

= ≠

= ≠

⎡ ⎤ ⎢ ⎥ ⎢ ⎥= + ⎢ ⎥− +⎢ ⎥ ⎣ ⎦ ⎡ ⎤⎛ ⎞ ⎢ ⎥+⎜ ⎟ ⎢ ⎥⎜ ⎟⋅ + +⎢ ⎥⎜ ⎟− +⎢ ⎥⎜ ⎟ ⎢ ⎥⎝ ⎠⎣ ⎦

∑

∑ ∑

∑

(29)

Given that 0 1ir≤ < , it can be followed that 1lim 1i i

r i

r r→

= ∞ −

,

and 1

i

i

r r−

is an increasing function of ri. As such, it can be

observed from (29) that K is also an increasing function of ri.

In addition, there must exists a 0r ( 2 1

2 1b ωσ =

+ ) such that

22 1

1 i

i

r b r

ωσ ≥ −

when 0ir r≥ . That is, K>1 holds when 0ir r≥ . Thus

it can be concluded from (28) that E[ ] 0i irπ∂ ∂ ≤ holds when 0ir r≥ .

For the case when 00 ir r≤ < , consider first that ri=0 and

rj=0(j≠i), the expression (29) can be rewritten as:

1,

1,

1, 1

(1 ) 1

1 1 1

n

k jn k k j

n n j j i

j k j j i k

b d d b K

b d b d

= ≠

= ≠

= ≠ =

⎡ ⎤ +⎢ ⎥

⎢ ⎥= ⋅ + ⎢ ⎥

+ +⎢ ⎥ ⎣ ⎦

∑ ∑

∑ ∑ (30)

Due to the fact that

1,

1, 1,

1

(1 )

1

n

k jn n k k j

jn j j i j j i

k k

b d d b b d

b d

= ≠

= ≠ = ≠

=

+ <

+

∑ ∑ ∑

∑ (31)

thus from (30), K<1 holds for ri=0 and rj=0(j≠i). From (29) it follows that K is a decreasing function of rj(j≠i). Therefore, K<1 holds for ri=0 and 0jr ≥ (j≠i), which means that

E[ ] 0i irπ∂ ∂ ≥ when ri=0 and 0jr ≥ (j≠i). K will increase with increasing ri. Since K is also affected by other generator’s risk preference rj(j≠i), it is possible that K becomes greater than 1 (which leads to E[ ] 0i irπ∂ ∂ ≤ ) when ri increases to a value smaller than r0.

Therefore, each generator’s expected profit will generally increase with increasing its risk aversion when its risk aversion is relatively low. For relatively high degrees of a generator’s risk aversion, the generator’s expected profit will decrease with increasing its risk aversion.

From (8), the first-order partial deviation of the variance of generator i’s profit with respect to its risk preference is:

2 2[ ] 2i iw i i i

Var Q b Q

r r π

σ ∂ ∂

= ⋅ ∂ ∂

(32)

Since 0i iQ r∂ ∂ ≤ , it is obvious that Var[ ] 0i irπ∂ ∂ ≤ , which means that the variance of generator’s profit will decrease with increasing its risk aversion.

These results can be explained to some extent according to the generator’s utility-maximization objective in (9). When the degree of a generator’s risk aversion is relatively low, the generator’s optimization objective is mainly to maximize its expected profit. As the degree of a generator’s risk aversion goes up, the weight of profit risk in the generator’s optimization objective will increase and the generator will sacrifice part of its expected profit to reduce its profit risk.

III. NUMERICAL EXAMPLES Consider a power market, the coefficients in the market

demand function in a certain time period (1h) are assumed to be: a=$80/MWh，b=$1.0/(MW)2 h. In this time period, the output of wind power units is assumed to follow a normal distribution with a mean of 10MW and a variance of 9.

We assume there are two symmetric conventional generators in the power market. The cost parameters of generator i(i=1,2) are: αi=$10.0/MWh, βi=$1.0/(MW)

2h. Impacts of the two generators’ risk preference on generator 1’s strategic variable, equilibrium output and the expected market price are illustrated in Fig. 1, 2 and 3, respectively.

Figure 1. Impact of risk preference on generator 1’s strategic variable

Figure 2. Impact of risk preference on generator 1’s output

Figure 3. Impact of risk preference on the expected market price

From Fig. 1, it can be found that generator 1’s strategic variable not only decreases with increasing its own risk aversion but also decreases with increasing its rivals’ risk aversion. However, the impact of its own risk aversion is much greater. In addition, the more risk neutral a generator is, the less impact its rivals’ risk preferences will have on its strategic variable. From Fig. 2, it can be observed that generator 1’s equilibrium output will decrease with increasing its risk aversion, and will increase with increasing its rivals’ risk aversion. However, the impact of its own risk aversion is much

greater. From Fig. 3, it can be seen that the expected market price will increase with increasing the degree of generators’ risk aversion.

The above results indicate that as a generator’s risk aversion increases, the generator will decrease its output by strategic bidding. This will lead to an increase in its rival’s output. However, the decrease in its own output is greater than the increase in its rival’s output, leading to an increase in the expected market price.

The impact of generators’ risk preferences on generator 1’s expected profit is depicted in Fig.4. It is shown that generator 1’s expected profit will generally increase with increasing its risk aversion when its risk aversion is relatively low. For relatively high degrees of generator 1’s risk aversion, its expected profit will decrease with increasing its risk aversion. This result is consistent with the theoretical analysis. In addition, generator 1’s expected profit increases with increasing its rivals’ risk aversion.

The impact of the generators’ risk preferences on the standard deviation of generator 1’s profit is given in Fig.5. The standard deviation of generator 1’s profit decreases with increasing its risk aversion, but slightly increases with increasing generator 2’s risk aversion. This means that increasing generator’s risk aversion can effectively reduce its profit risk, while an increase in its rivals’ risk aversion will have an adverse effect on its profit risk.

Figure 4. Impact of risk preference on generator 1’s expected profit

Figure 5. Impact of risk preference on the standard deviation of generator

1’s profit

IV. CONCLUSIONS In this paper, a supply function equilibrium model of

electricity markets with wind power is developed taking into account the conventional strategic generators’ risk preference. The impacts of generator’s risk preference on the equilibrium results are examined by theoretical analysis and numerical simulation. It is shown that as the degree of a generator’s risk aversion increases, the generator will decrease its output by strategic bidding. This will lead to an increase in its rival’s output. However, the decrease in its own output is greater than the increase in its rival’s output, leading to an increase in the expected market price. Furthermore, it is also demonstrated that each generator’s expected profit will generally increase with increasing its risk aversion when its risk aversion is relatively low. For relatively high degrees of a generator’s risk aversion, its expected profit will decrease with increasing its risk aversion. Each generator’s expected profit will increase with increasing its rivals’ risk aversion. In addition, increase in the degree of a generator’s risk aversion can effectively reduce its profit risk, but increasing its rivals’ risk aversion will have an adverse effect on its profit risk.

REFERENCES [1] U. Helman, “Market power monitoring and mitigation in the US

wholesale power markets,” Energy, vol. 31, no. 6-7, pp. 877-904, May-June 2006.

[2] M. Ventosa, A. Bayllo, A. Ramos, and M. Rivier, “Electricity market modeling trends,” Energy Policy, vol. 33, no. 7, pp. 897-913, May 2005.

[3] J. Smith, “Wind power: present realities and future possibilities,” Proc. IEEE, vol. 97, no. 2, pp. 195-197, Feb. 2009.

[4] J. Xu, D. He, and X. Zhao, “Status and prospects of Chinese wind energy,” Energy, vol. 35, no. 11, pp. 4439-4444, Nov. 2010.

[5] A. Botterud, J. Wang, V. Miranda, and R. J. Bessa, “Wind power forecasting in U.S. electricity markets,” Elect. J., vol. 23, no. 3, pp. 71-82, Apr. 2010.

[6] D. Feng, D. Gan, J. Zhong, and Y. Ni, “Supplier asset allocation in a pool-based electricity market,” IEEE Trans. on Power Syst., vol. 22, no. 3, pp 1129-1138, Aug. 2007.

[7] M. Liu, F. Wu, “Managing price risk in a multimarket environment,” IEEE Trans. on Power Syst., vol. 21, no. 4, pp. 1512-1519, Nov. 2006.

[8] A. Conejo, M. Carrion, “Risk-constrained electricity procurement for a large consumer,” IET Gen., Transm., Distrib., vol. 153, no. 4, pp. 407-413, July 2006.

[9] D. Feng, D. Gan, and J. Zhong, “Supplier asset allocation in a Pool-based electricity market,” IEEE Trans. on Power Syst., vol. 22, no. 3, pp. 1129-1138,Aug. 2007.

[10] F. Azevedo, Z. Vale, and P. Oliveira, “A decision-support system based on particle swarm optimization for multi-period hedging in electricity markets,” IEEE Trans. on Power Syst., vol. 22, no. 3, pp. 995-1103, Aug. 2007.

[11] Y. Liu, F. Wu, “Risk management of generators’ strategic bidding in dynamic oligopolistic electricity market using optimal control,” IET Gen., Transm., Distrib., vol. 1, no. 3, pp. 388–398, May, 2007.

[12] X. Guan, J. Wu, and F. Gao, “Optimization-based generation asset allocation for forward and spot market,” IEEE Trans. on Power Syst., vol. 23, no. 4, pp. 1796-1808, Nov. 2008.

[13] M. Carrion, A. Conejo, and J. Arroyo, “Forward contracting and selling price determination for a retailer,” IEEE Trans. on Power Syst., vol. 22, no. 4, pp. 2105-2114, Nov. 2007.

[14] A. Conejo, G. Raquel, and M. Carrion, “Optimal involvement in futures markets of a power producer,” IEEE Trans. on Power Syst., vol. 23, no. 2, pp. 703-711, May 2008.

[15] R. Dahlgren, C. Liu, and L. Lawarrée, “Risk assessment in energy trading,” IEEE Trans. on Power Syst., vol. 18, no. 2, pp. 503-511, May 2003.

[16] Z. Yuan, D. Liu, C. Jiang, and Z. Hou, “Analysis of supplier

equilibrium strategy considering transmission constraints,” IET Gen., Transm., Distrib., vol. 152, no. 1, pp. 17-22, Jan. 2005.

[17] K. Neuhoff, J. Barquin, M. Boots, A. Ehrenmann, B. Hobbs, and F. Rijkers, et al, “Network-constrained Cournot models of liberalized electricity markets: The devil is in the details,” Energy Econ., vol. 27, no. 3, pp. 495-525, May 2005.

[18] H. Chen, X. Wang, “Strategic behavior and equilibrium in experimental oligopolistic electricity markets,” IEEE Trans. on Power Syst., vol. 22, no. 4, pp. 1707-1716, Nov. 2007.

[19] H. Chen, K. Wong, H. Nguyen, and C. Y. Chung, “Analyzing oligopolistic electricity market using coevolutionary computation,” IEEE Trans. on Power Syst., vol. 21, no. 1, pp. 143-152, Feb. 2006.

[20] U. Helman, B. Hobbs, “Large-scale market power modeling: analysis of the U.S. eastern interconnection and regulatory applications,” IEEE Trans. on Power Syst., vol. 25, no. 3, pp. 1434-1448, Aug. 2010.

[21] X. Wang, Y. Li, S. Zhang, “Oligopolistic equilibrium analysis for electricity markets: a nonlinear complementarity approach,” IEEE Trans. on Power Syst., vol. 19, no. 3, pp. 1348-1355, Aug. 2004.

[22] H. Niu, R. Baldick, G. Zhu, “Supply function equilibrium bidding strategies with fixed forward contracts,” IEEE Trans. on Power Syst., vol. 20, no. 4, pp. 1859-1867, Nov. 2005.

[23] H. Chen, K. Wong, C. Chung, and H. Nguyen, “A coevolutionary approach to analyzing supply function equilibrium mode,” IEEE Trans. on Power Syst., vol. 21, no. 3, pp. 1019-1028, Aug. 2006.

[24] Y. Liu, F. Wu, “Impacts of network constraints on electricity market equilibrium,” IEEE Trans. on Power Syst., vol. 22, no. 1, pp. 126-134, Feb. 2007.

[25] E. J. Anderson, X. Hu, “Forward contracts and market power in an electricity market,” Int. J. Ind. Org., vol. 26, no. 3, pp. 679-694, May 2008.

[26] C. Yu, S. Zhang, X. Wang, and T. S. Chung, “Modeling and analysis of strategic forward contracting in transmission constrained power markets,” Electr. Power Syst. Res., vol. 80, no. 3, pp. 354-361, Mar. 2010.

Selling Wind Power in Electricity Markets: The status today, the opportunities tomorrow

Eilyan Y. Bitar & Kameshwar Poolla

Abstract— California has set a target of 33% penetration from all renewable sources by 2020. Wind energy must play a key role in realizing such aggressive targets. At these deep penetration levels, integration of utility scale wind production into the electricity grid poses serious engineering and market challenges. These are due to the uncertainty, intermittency, and uncontrollability of wind power. Wind is random.

Today, wind energy is assimilated into the grid by legislative mandates, feed-in tariffs, favorable penalty pricing, guaranteed grid access, and/or construction subsidies. The variability in production is absorbed by scheduling operating reserves. For example, in California, the Participating Intermittent Resource Program (PIRP) legislation compels the system operator to accept all produced wind power subject to certain contractual constraints. This amounts to a system take-all-wind scenario in which wind power is treated as a negative load. The burden of reserve costs is socialized among the load serving entities (LSE).

This extra-market approach works at today’s modest pene- tration levels, but will become untenable as wind penetration increases for both economic and environmental reasons. We discuss the consequences on wind integration in the near term as wind power producers are forced to participate in competitive electricity markets alongside conventional dispatch- able generation. We explore aggregation, firming strategies, and fair reserve cost allocation. In the long term, we argue that new market mechanisms are required to deal with wind power variability. These include intra-day markets to leverage on improved forecast accuracy on shorter horizons, bilateral contracts with interruptible loads such as electric vehicles, and most radically, the possibility of selling random power through price differentiated quality of supply.

I. THE VARIABILITY CHALLENGE

Wind power is inherently variable [12]. It is non- dispatchable, highly intermittent, and difficult to forecast on horizons beyond five minutes. At levels of deep renewable penetration, these generation variability characteristics will pose formidable challenges to the preservation of instanta- neous balance between supply and demand, while simulta- neously respecting system security constraints.

In contrast to variable renewable generators, the majority of conventional thermal generators in use today are funda- mentally dispatchable and, to a large extent, predictable. Nonetheless, the power system architecture and attending operations have been designed to explicitly deal with the

Supported in part by the NSF under Grants EECS-0925337, ECCS- 1129001. Thanks to Pramod Khargonekar, Pravin Varaiya, and Ram Ra- jagopal for many helpful discussions.

Corresponding author: E.Y. Bitar is with the Department of Electrical and Computer Engineering, Cornell University [email protected]

K. Poolla is with the Department of Electrical Engineering and Computer Science, U.C. Berkeley [email protected]

variability emanating from natural fluctuations in load and unplanned contingency events such as branch and/or gen- eration outages. Generally speaking, imbalances arising be- tween generation and load are compensated for by reserve generation capacity procured by the independent system operator (ISO) through ancillary services (AS) markets. The various phenomena responsible for system imbalances occur on differing time scales and thus require the procurement of reserve resources with a variety of response capabilities [11], [13]. The subsequent cost of the procured reserve capacity is then socialized among the participating load serving entities (LSE) based on their relative contribution to the total demand. To a large extent, the current approach to compensating variability amounts to a paradigm in which generation is tailored to follow load. Given the relative success of the status quo, it is tempting to presume that the added variability of renewable generation can be similarly compensated for with existing reserve mechanisms. In fact, this is the approach taken by many balancing authorities within the United States: all wind and solar power pro- duction is taken by the system operator and the attendant variability is compensated for with existing reserve margins. As we will see in the following section, this approach to renewable energy integration will become untenable at deep penetration levels.

II. DEALING WITH VARIABILITY TODAY

Today, wind and solar energy are assimilated into the grid through legislative mandates, feed-in tariffs, lenient penalty pricing, guaranteed grid access, and/or construc- tion subsidies. Specifically, in California, the Participating Intermittent Resource Program (PIRP) legislation compels the independent system operator (ISO) to accept all produced wind power subject to certain contractual constraints. This amounts to a system take-all-wind scenario in which wind power is treated as a negative load and the subsequent increase in the variability of net-load is absorbed by a port- folio of reserve generation capacity, whose cost is allocated among the load serving entities (LSE). This socialization of added reserve costs among the LSEs can be interpreted as an implicit subsidy for variability costs to participating wind power producers. Accordingly, there are ongoing public policy and operational procedure debates regarding the fair allocation of the costs of these increased reserves [15].

Even at today’s modest levels of penetration, the added variability due to wind results in systemic operational prob- lems. For example, on February 26, 2008, the Electric

2012 American Control Conference Fairmont Queen Elizabeth, Montréal, Canada June 27-June 29, 2012

978-1-4577-1096-4/12/$26.00 ©2012 AACC 3144

Reliability Council of Texas (ERCOT) had to declare an emergency load curtailment plan due in part to an inaccurate forecast of wind power production [6]. The impact of inter- mittency and inaccurate forecasting on reserve margins will only become more pronounced as wind energy penetration increases [9], [8], [7]. In order to quantify this statement, several wind integration studies have computed detailed estimates of the increase in reserve requirements needed to compensate the added variability due to wind under a system- take-all wind regime. For example, the 2010 EWITS report [7] by NREL projects that regulating reserve requirements will increase by 1500 MW (on average) under a 20% percent wind energy penetration scenario in the PJM interconnection. Such an increase in reserve requirements is unacceptable. It is too expensive. Ergo, it will rapidly become infeasible to continue the implicit subsidization of the variability costs among the load serving entities. Moreover, it severely mit- igates the net greenhouse gas benefit of renewable energy, as regulating reserves are normally supplied by fast-acting, fossil fuel based thermal generators such as natural gas turbines. Clearly, the current strategy cannot scale.

III. THE OPPORTUNITIES TOMORROW

As renewable energy penetration increases, how must the assimilation of variable power evolve, so as to minimize integration costs, while maximizing the net environmental benefit? Clearly, strategies that mitigate additional reserve requirements will be an essential means to this end. Such strategies will fundamentally fall into three (overlapping) categories: (a) direct reduction of variability in generation, (b) new market mechanisms to absorb variability, and (c) demand-side solutions that use flexibility to adapt to vari- ability in generation.

A. Generation-side Firming Solutions

In the near term, it is likely that wind and solar power producers will be faced with increased exposure to market signals that incentivize reduction in output variability – a stark contrast to the California Participating Intermittent Resource Program (PIRP). For example, in the United King- dom, large wind power producers are forced to participate in conventional wholesale electricity markets where they are subject to ex-post financial penalties for deviations from contracts offered ex-ante in forward markets [1], [3] – thus eliminating the implicit subsidy for variability costs. The implementation of imbalance penalty mechanisms represents an initial departure from the system-take-all-wind approach. In response to the financial risk emanating from uncertainty in wind power production, a rational wind power producer will be forced to curtail its projected output, thus decreasing the amount of variability that has to be compensated for with reserve generation by the system operator. However, such a removal of the implicit subsidy for variability cost may result in significant profit loss to the wind power producer. Consequently, it will become necessary for the wind power producer to develop and evaluate strategies that aid in the mitigation of generation variability. We refer to this as

firming of variable renewable power. Potential approaches to firming include: Improved Forecasting [20], [21], [22], Storage [25], [24], Renewable Resource Aggregation [19], Local Generation [3].

A fundamental question arises in this setting: What is the marginal value of a given firming strategy? For example, what is the subsequent increase in expected profit to a renewable power producer per MWh of co-located storage capacity [2]? What is the sensitivity of expected profit to forecast uncertainty? Such a monetization of the aforemen- tioned firming strategies will play a central role in shaping investment decisions.

B. Market Mechanism Solutions Firming mechanisms alone, may offer substantial near-

term benefits. However, if we are to transition to a power system with >30% of the energy supply coming from variable renewable resources, we will have to fundamentally rethink the way the electricity grid is operated. In the long- term, we argue that new market systems and instruments to explicitly address the difficulties with variable generation are required. For example, wind power forecast accuracy tends to steadily decrease with the shortening of the prediction horizon below five hours. Hence, the creation of intra-day en- ergy and ancillary service markets (spanning the day-ahead and real-time markets) [18] will introduce additional trading opportunities to leverage on improved forecast accuracy on shorter horizons. In this manner, renewable power producers will have the opportunity to incrementally offer their variable energy in a sequence of intra-day markets – allowing for a more effective balance of risk and return.

More radically, we envision that market systems will evolve to allow for price differentiated quality of supply. Traditionally, the electric grid has been operated such that generation is tailored to counteract the variability in load. Load is largely treated as inelastic. However, there exists significant flexibility inherent to load that is currently not utilized. The power requirements of many commercial and residential loads are such that a fraction of instantaneous power demand at any given moment is inherently deferrable in time (subject to certain deadlines on delivery). Examples include thermal systems such as refrigerators, water heaters, HVAC systems, and, assuming mass adoption of plug-in electric vehicles, batteries. Given the inherent deferrability of such loads, certain customers may be willing to accept varying degrees of interruptible power supply in exchange for a lower price, all without experiencing significant loss of utility. Clearly, there is an opportunity and need to design novel market systems that provide flexible consumers the op- tion to purchase multiple quantities of energy – with varying degrees of reliability – from variable generators. Initial work, along these lines, on efficient pricing of interruptible power service contracts can be found in Tan et al. [14]. Further, in addition to participating in such novel market systems, flexible load devices – if intelligently controlled in aggregate – may be capable of providing various ancillary services (e.g., regulation, load-following) to the system operator at a much lower cost than conventional reserve generation, [4].

3145

As the penetration of variable renewable generation con- tinues to increase, we must necessarily transition to a modus operandi in which load is elevated to the equivalent status of dispatchable generation. In this way, load can be tailored to absorb variability in supply.

C. Coordinated Demand-side Solutions

On the demand side, the primary instrument to deal with variability of renewable generation is coordinated aggrega- tion of networked resources at the distribution side including renewable and micro-generation, electricity storage, smart appliances, and responsive loads. Coordinated Aggregation [16], [17] is substantially more powerful than traditional demand response (DR) or naive aggregation. It (a) involves the intelligent control of deferrable loads and available storage to match the variable generation, (b) requires Smart Grid communication and computation infrastructure, and (c) creates bulk power and ancillary service market opportunities that are much more substantial that those offered by the peak- shaving capacity markets. Existing DR research has focused on the limited market opportunities of (a) shifting load from peak [26], [27], [28], (b) contingency reserves, [29], [32], [30], and (c) decentralized frequency response [31], [34], [33]. Simple aggregation appeals to the fact that the sum of random individual demands or renewable resources will frequently have a smaller variance than the sum of the individual variances. The value of coordinated aggregation is derived from the improved forecastability it offers in ex ante markets. This, in turn, will require less capacity reserve requirements for load following. There are two approaches to coordination that have been studied in the literature: (a) direct-load control [29], where resources are centrally managed, and (b) indirect load-control [35], where resources respond to generation conditions through the proxy of real- time prices. We submit that direct-load control is the more attractive option as it offers greater flexibility and would face fewer adoption obstacles as customers are not exposed to real-time price volatility.

A central challenge in using coordinated aggregation to absorb variability from renewable generation is that of re- warding resource participation. This has two components: (a) the avoided cost benefits of requiring smaller operating reserves must be equitably monetized, and (b) resources must be rewarded for participation. Harvesting revenue from absorbing variability will require mandating new ancillary service markets. This expensive proposition faces regula- tory hurdles, fair pricing mechanisms, and acceptance by renewable generators. Incentivizing participating resources is difficult because the rewards are so small. Options that we will discuss include lottery mechanisms and up-front adoption rebates.

REFERENCES

[1] J. R. Abbad, “Electricity market participation of wind farms: the success story of the Spanish pragmatism,” Energy Policy, vol. 38, no. 7, pp. 3174-3179, Jul 2010.

[2] E.Y. Bitar, R. Rajagopal, P.P. Khargonekar, K. Poolla, “The Role of Co-located Storage for Wind Power in Conventional Electricity Mar- kets,” Proceedings of the 30th IEEE American Control Conference, San Francisco, CA, 2011.

[3] E. Bitar et al., “Bringing Wind Energy to Market,” To Appear, IEEE Transactions on Power Systems, 2012.

[4] D.S. Callaway and I.A. Hiskens, “Achieving Controllability of Electric Loads,” accepted for publication in Proceedings of the IEEE, 2011.

[5] Committee on Stabilization Targets for Atmospheric Greenhouse Gas Concentrations; National Research Council, “Climate Stabilization Targets: Emissions, Concentrations, and Impacts over Decades to Millennia,” The National Academies Press, Washington, D.C., USA, 2011.

[6] E. Ela and B. Kirby, “ERCOT event on February 26, 2008: Lessons learned,” NREL Technical Report, NREL/TP-500-43373, July 2008.

[7] EnerNex Corp., Eastern Wind Integration and Transmission Study, National Renewable Energy Laboratory, Report NREL/SR-550-47078, January 2010.

[8] GE Energy, Western Wind and Solar Integration Study, National Renewable Energy Laboratory, Report NREL/SR-550-47434, May 2010.

[9] GE Energy, “Analysis of Wind Generation Impact on ERCOT An- cillary Services Requirements,” Prepared for ERCOT, Final Report, March 28, 2008.

[10] H. Holttinen et al., “Impacts of large amounts of wind power on design and operation of power systems, results of IEA collaboration,” 8th International Workshop on LargeScale Integration of Wind Power into Power Systems, 14-15 Oct. 2009 Bremen.

[11] North American Electric Reliability Corporation, “Reliability Stan- dards for the Bulk Electric Systems of North America,” November 2009.

[12] North American Electric Reliability Corporation (NERC), “Accommo- dating High Levels of Variable Generation,” Special Report, Princeton, NJ, USA, April, 2009.

[13] Y. Rebours and D. Kirschen, “A survey of definitions and specifications of reserve services,” Dept. of Electrical Engineering, The University of Manchester, October 2005.

[14] C.W. Tan, and P.P. Varaiya, “Interruptible electric power service contracts,” Journal of Economic Dynamics and Control 17 (May, 1993): 495-517.

[15] L. Vandezandea, L. Meeusa, R. Belmansa, M. Saguanb and Jean- Michel Glachant, “Well-functioning balancing markets: A prerequisite for wind power integration,” Energy Policy, Vol. 38, pp. 3146:3154, July 2010.

[16] D. Bakken et al., “GRIP - Grids with Intelligent Periphery: Control Architectures for Grid2050,” IEEE SmartGridComm, Architectures and Models for the Smart Grid, 2011.

[17] A. Subramanian et al., “Real-time Scheduling of Deferrable Electric Loads,” To Appear, Proceedings of the Automatic Control Conference, 2012.

[18] P.P. Varaiya, F. Wu, J.W. Bialek, “Smart Operation of Smart Grid: Risk-limiting Dispatch,” Proceedings of the IEEE , vol.99, no.1, pp.40- 57, Jan. 2011

[19] E. Baeyens et al., “Wind energy aggregation: a coalitional game approach,” To Appear, Proceedings of the Conf. on Decision and Control, Orlando, FL, 2011.

[20] J. Pease, “Critical short-term forecasting needs for large and unsched- uled wind energy on the BPA system,” 3rd Workshop on the Best Practice in the Use of Short-term Forecasting of Wind Power, October 2009.

[21] P. Pinson, H. A. Nielsen,H. Madsen, and K. Kariniotakis, “Skill Forecasting from Ensemble Predictions of Wind Power,” Applied Energy, vol. 86, pp. 1326-1334, 2009.

[22] A. Botterud, J. Wang, C. Monteiro, and V. Miranda, “Wind Power Forecasting and Electricity Market Operations,” Proceedings of US- AEE 2009.

[23] E.D. Castronuovoa and J.A. Pecas Lopes, “Optimal operation and hydro storage sizing of wind-hydro power plant,” Journal of Electrical Power and Energy Systems, 26 (2004), 771-778.

[24] A. Cavallo, “Controllable and affordable utility-scale electricity from intermittent wind resources and compressed air energy storage (CAES),” Energy, 2007;32(2):1207.

[25] P. Denholm, E. Ela, B. Kirby, and M. Milligan, “The Role of Energy Storage with Renewable Electricity Generation,” Technical Report NREL/TP-6A2-47187, January 2010.

3146

[26] T.F. Lee, M.Y. Cho, Y.C. Hsiao, P.J. Chao, and F.M. Fang, “Opti- mization and implementation of a load control scheduler using relaxed dynamic programming for large air conditioner loads,” IEEE Trans. Power Syst., vol. 23, no.2, pp.691-702, May 2008.

[27] L. Yao and H.R. Lu, “A two-way direct control of central air- conditioning load via the internet,” IEEE Trans. Power Del., vol. 24, no. 1, pp. 240-248, January 2009.

[28] L. Jiang and S. Low, “Multi-period optimal procurement and demand responses in the presence of uncertain supply,” Proc. of the 50th Conf. on Decision and Control, December 2011.

[29] L. Goel, Q. Wu, and P. Wang, “Fuzzy logic-based direct load control of air conditioning loads considering nodal reliability characteristics in restructured power systems,” Electric Power Systems Research, vol. 80, no.1, pp. 98-107, January 2010.

[30] G. Heffner, C. Goldman, B. Kirby, and M. Kintner-Meyer, “Loads providing ancillary services: Review of international experience,” Technical Report LBNL-62701, Lawrence Berkeley National Labo- ratory, May 2007.

[31] J.A. Short, D.G. Infield, and L.L. Freris, “Stabilization of grid fre- quency through dynamic demand control,” IEEE Trans. Power Syst., vol. 22, no. 3, pp.1284-1293, August 2007.

[32] J.H. Eto, J. Nelson-Hoffman, C. Torres, et al., “Demand response spinning reserve demonstration,” Technical Report LBNL-62761, Lawrence Berkeley National Laboratory, May 2007.

[33] D.J. Hammerstrom, J. Brous, D.P. Chassin, et al., “Pacific Northwest GridWise Testbed Demonstration Projects,” Tech. Rep. PNNL-17167, Pacific Northwest National Laboratory, October 2007.

[34] U.K. Market Transformation Program, “Dynamic demand control of domestic appliances,” Technical report, Market Transformation Programme, 2008.

[35] D.S. Callaway, “Tapping the energy storage potential in electric loads to deliver load following and regulation, with application to wind energy,” Energy Conversion and Management, vol. 50, no. 5, pp.1389- 1400, May 2009.

3147