Increasing Wind Capacity Value in Tasmania Using Wind and Hydro Power Coordination

Mehdi Mosadeghy, Student Member, IEEE, Tapan Kumar Saha, Senior Member, IEEE and Ruifeng Yan, Member, IEEE

School of Information Technology & Electrical Engineering The University of Queensland Brisbane, Australia [email protected], [email protected], [email protected]

Abstract—Wind power penetration is increasing in power system adequacy assessment [4]-[6], and some others have considered in many countries. To mitigate wind power fluctuations and wind power as a negative load [7]. It has been shown that the maximize clean energy utilization, coordinating wind and hydro impacts of wind power on the reliability of power system power is proposed in the literature. This paper investigates the depends on many factors such as correlation between wind benefits of coordinating hydro generators with wind farms, from speed and system load, wind regime, wind power penetration reliability point of view in Tasmania, Australia. Reliability level, etc. [8]. indices are computed in generation system adequacy assessment using the state sampling Monte Carlo technique. In accordance Tasmania is an island state of Australia, where the with these indices the capacity value of Tasmania’s wind power majority of its power production is from renewable energy is calculated and the impact of hydro power coordination on this resources. Power generation in Tasmania is currently almost value is demonstrated. The results show that using hydro units entirely produced by hydro generation with a total installed to support wind power will improve the capacity value of the capacity of 2,300MW [1]. Installed wind generation capacity wind farms regarding their wind regime and the combined is already around 140 MW with a further 500MW under capacity. construction or publicly announced for future development Tasmania has the potential for high wind penetration levels Index Terms— Capacity value, Reliability evaluation, [9]. Coordinating the output of some existing hydro units with Tasmanian power system, Wind and hydro generation wind farms can balance the wind power fluctuations and coordination. increase the benefits of wind farms. Power system reliability in the presence of wind-hydro coordination has been studied in I. INTRODUCTION [10] and [11]. Hydro units were considered as peak time Implementation of policies such as Large-scale Renewable generators in [10] and it was shown that coordinating wind Energy Target (LRET) and carbon pricing in Australia will be and hydro can improve system’s adequacy when a proper important regulatory drivers for renewable generation over the number of hydro units are designated to support wind farms. next 10 years. Based on current renewable energy costs, wind Wind power is utilized in [11] to store water in hydro power could be the main renewable generation technology [1]. reservoirs during off peak periods, which can be used at a later However, the intermittency of wind means that integrating time by the hydro station to produce its highest capacity large scale wind power will affect many aspects of power during peak load times when electricity has a higher value. system. Evaluating the contribution of wind generators in the From a reliability perspective, significant work has been system adequacy of the Tasmanian power system is essential undertaken to study the effects of wind power integration on due to the expected increase in the wind power penetration system reliability using analytic and probabilistic methods level in Tasmania. This paper investigates the capacity value [2]–[8]. Reliability indices for different wind speed models of wind power in Tasmania using the Woolnorth and such as Markov Chain, Normal distribution and autoregressive Musselroe wind farms which have different wind regimes. moving average (ARMA) time series were compared in [2] Wind power is modeled as a negative load and the state using Monte Carlo simulation and analytical method. Some sampling Monte Carlo simulation method is applied to population-based intelligent (PIS) procedures were applied in calculate reliability indices for the Tasmanian power system. [3] to search the system states and obtain the reliability indices This work also explains the impact of hydro power in the presence of wind generators. Some studies have adopted coordination with different allocated capacities on the wind different multistate models for wind power in system power capacity value in Tasmania.

978-1-4799-1303-9/13/$31.00 ©2013 IEEE The Tasmanian power system is briefly described in The statistical values of wind speed data from 2006 to Section II. Section III describes the model of coordinating 2012 for these sites provided by Australian Bureau of wind and hydro power. In Section IV, the wind capacity value Meteorology are presented in Table I. Fig. 2 illustrates the is calculated for Tasmanian wind farms and the impact of wind speed probability distributions for Woolnorth and wind–hydro coordination on Tasmanian wind capacity value is Musselroe sites during winter months. evaluated. Finally, conclusions are summarized in Section V.

II. TASMANIAN POWER SYSTEM A. General Information Tasmania has a total installed generating capacity of 2,800 MW and is connected to mainland Australia via a single HVDC undersea link with an export capacity of 630 MW. Tasmania generates most of its electricity from hydro with total installed capacity around 2,300MW and also has around 386MW of gas-fired generation. [1].

Figure 2. Winter wind speed probability distibution for Woolnorth and Musselroe sites from 2007 to 2012

III. COORDINATING WIND AND HYDRO UNITS Hydroelectric power is the major source of electricity in Tasmania with 27 hydro power stations, generating around 10,000 GWh of electricity annually [9]. This presents an ideal opportunity for the Tasmanian power system to coordinate hydro-electric and wind power generation to balance wind power variations and maximize the renewable energy Figure 1. Mean value of Tasmania hourly winter load for over 6 years utilization. Among Tasmanian hydro stations, the Gordon (2007-2012) scheme located in the southern part of Tasmania is considered to be ideal for coordination with wind power. The Gordon Tasmania peak load generally occurs during winter. station has the highest hydro installed capacity in Tasmania Fig.1shows the average value of Tasmania hourly load in [1] comprising three 144MW units supplied with water from June, July and August for over 6 years from 2007 to 2012 Lake Gordon [14]. Fig. 3 displays the location of Woolnorth [12]. It can be seen that the hourly load profile consists of two and Musselroe wind sites and the Gordon hydro station in peak periods and the average load during July is higher than Tasmania. other months.

B. Wind Data Tasmania is geographically located within the 'Roaring Forties'; latitudes with some of the most reliable winds on earth [9]. Tasmania currently has around 140 MW of wind power installed at the Woolnorth wind farm. Construction of the 168 MW Musselroe wind farm is expected to be completed by June 2013[1]. A further 330 MW of new wind power proposals are publicly announced [13].

TABLE I. WIND SPEED STATISTICAL DATA OF 6 YEARS FOR WOOLNORTH AND MUSSELROE SITES

Site Mean (m/s) Standard Deviation (m/s)

Woolnorth 9.73 4.94 Musselroe 6.80 3.65 Figure 3. Locations of Gordon hydro station and Tasmanian wind farms

In this paper the Gordon hydro station is coordinated with Tasmanian wind farms in a daily energy balancing policy. During off peak periods the wind farm produces power and is described as in Fig.5.The first step requires the system load the hydro units decrease electricity production which to be clustered into 20 levels using the K-means method [17]. effectively stores water in its reservoir. During peak load Table II represents the 20-step load levels for the Tasmanian periods, the Gordon hydro power station could generate power system during winter from 2007 to 2010. The load additional power using the stored water to maximize its output levels are in per unit of the 1,800 MW peak load. when electricity usually has a higher value. However, some factors such as wind regime, load profile and storage constraints need to be considered to coordinate wind and hydro generation [11]. Fig. 4 shows the water level in the Gordon hydro station reservoir during 2010 and 2011 [15]. It can be seen that the reservoir typically drops to lower levels during winter.

Figure 5. Capacity value of Woolnorth wind farm

This clustered load and the state sampling Monte Carlo method are utilized to obtain the LOEE index for the Tasmanian power system without the wind farm. The dashed Figure 4. Gordon Reservoir levels in 2010 and 2011[15] line in Fig. 5 illustrates the loss of energy expectation for different extra loads added to the power system. It can be seen IV. CAPACITY VALUE OF WIND FARMS that LOEE is around 200 MWh/yr. for 1,800 MW peak load (no extra load). The capacity value is generally defined as the effective load carrying capability. This value represents the amount of TABLE II. THE 20-STEP LOAD LEVEL MODEL FOR TASMANIAN load that can be added to power system at a defined reliability SYSTEM WITHOUT WIND POWER level in presence of a new generator [16]. The calculation Step Load Level Step Load Level Probability Probability method and required data are described in [7] and [16]. No. (p.u.) No. (p.u.) In this paper loss of energy expectation (LOEE) is defined 1 0.5224 0.0164 11 0.7291 0.0674 as the reliability index to determine the capacity value of wind 2 0.5499 0.0314 12 0.7484 0.0734 farms in Tasmania. LOEE is the expected energy not supplied 3 0.5734 0.0417 13 0.7680 0.0707 while load exceeds the available generating capacity. Severity 4 0.5939 0.0475 14 0.7885 0.0683 of deficiencies, the number of occurrences and their duration 5 0.6131 0.0567 15 0.8109 0.0623 are included in LOEE [17]. 6 0.6319 0.0595 16 0.8350 0.0573 Available capacity of generators can be obtained regarding 7 0.6501 0.0630 17 0.8616 0.0484 their state. Demand not supplied (DNS) due to the load level 8 0.6696 0.0611 18 0.8918 0.0340 D exceeding the available generating capacity can be 9 0.6894 0.0607 19 0.9344 0.0151 computed by (1): 10 0.7092 0.0648 20 0.99 0.0002

m Wind farm output power is obtained using half-hourly DNSmax 0, D G (1) k ik wind speed data and wind turbine power curve [18]. A 0.04 i1 forced outage rate (FOR) is considered for wind turbines. The Where Gi denotes the available capacity of the ith wind farm output power is taken into account as negative load generator in the kth sampling. The annualized loss of energy and the equivalent load is clustered. Load levels for Tasmania expectation for N samples can be obtained using (2): with Woolnorth wind farm are given in Table III.

N The expected energy not supplied is computed once again  DNSk 8760 based on new clustered load levels. The solid line in Fig. 5 LOEE  k 1 (2) depicts the LOEE of this system. Comparing these two lines, a N 44 MW load could be added to Tasmanian power system supplemented by the Woolnorth wind farm without exceeding A. Without Hydro Coordination the existing LOEE level (200 MWh/yr.). This is equivalent to 1) Existing Woolnorth Wind Farm approximately 31% of the 140MW installed capacity of the The method used to calculate the capacity value of the Woolnorth wind farm. Woolnorth wind farm without considering hydro coordination TABLE III. THE 20-STEP LOAD LEVEL MODEL FOR TASMANIAN B. Wind – Hydro Coordination SYSTEM WITH WOOLNOTH WIND FARM In this section, hydro power is coordinated with wind Step Load Level Step Load Level Probability Probability power to increase their capacity value. Different values can be No. (p.u.) No. (p.u.) selected as the coordination criterion, which means the 1 0.4563 0.0189 11 0.6945 0.0665 minimum required capacity of the wind to start preserving 2 0.4943 0.0326 12 0.7146 0.0660 water in storage [11]. In this study the unused capacity of 3 0.5238 0.0386 13 0.7359 0.0703 Gordon station during peak time, which is around 30 MW, is 4 0.5481 0.0447 14 0.7584 0.0684 utilized as an inception strategy. During off peak times, when 5 0.5707 0.0540 15 0.7828 0.0609 the wind farm power exceeds 30 MW, Gordon hydro unit 6 0.5930 0.0579 16 0.8096 0.0542 stops producing power and retains its water level. This stored 7 0.6143 0.0638 17 0.8396 0.0475 water can be used to obtain maximum power from this hydro 8 0.6344 0.0716 18 0.8744 0.0344 station if it’s required. 9 0.6544 0.0673 19 0.9207 0.0144 The 20-step clustered load levels of Tasmania considering 10 0.6741 0.0678 20 0.9779 0.0002 this coordination for Woolnorth wind farm are presented in Table IV. This clustered load indicates that the probability of 2) Considering Future Musselroe Wind Farm higher load levels has reduced and the effective load after coordination occurs in medium load levels with more To evaluate the capacity value of the Musselroe wind possibility. farm, which will be operational by 2013, the autoregressive moving average (ARMA) model in (3) is applied to simulate TABLE IV. THE 20-STEP LOAD LEVELS CONSIDERING COORDINATION wind speed [19]. BETWEEN WOOLNOTH WIND FARM AND GORDON HYDRO STATION Step Load Level Step Load Level nn1 Probability Probability No. (p.u.) No. (p.u.) yyt i t i   t   j  t j (3) ij11 1 0.4561 0.0077 11 0.6865 0.0762 2 0.4897 0.0228 12 0.7086 0.0713 Where yt represents the data series value, 휙 is the auto- 3 0.5189 0.0328 13 0.7304 0.0767 regressive parameter and θ denotes the moving average 4 0.5438 0.0422 14 0.7532 0.0696 element. 훼 indicates white noise with a normal independent t 5 0.5657 0.0501 15 0.7772 0.0647 distribution (NID). 6 0.5869 0.0584 16 0.8043 0.0554 yt can be calculated using (4), where OWt is the observed 7 0.6070 0.0625 17 0.8343 0.0467 wind speed data for a site near Musselroe wind farm while μt 8 0.6263 0.0657 18 0.8697 0.0311 is the hourly mean and σt denotes the standard deviation of 9 0.6447 0.0769 19 0.9168 0.0137 these wind speed data. 10 0.6648 0.0754 20 0.9875 0.0002 y()/ OW  (4) t t t t This load has been utilized to calculate the LOEE index of The half-hourly wind data during June, July and August Tasmanian power system. The method described in the from 2007 to 2012 is utilized to build the ARMA time series previous subsection is applied to evaluate the effective load model for Musselroe wind site. Parameters of ARMA(4,3) carrying capability of wind farms coordinated with Gordon model of this wind regime are given in (5): hydro station. The effect of this coordination on the capacity value of Woolnorth and Musselroe wind farms is displayed in yt1.0148 y t1 0.7822 y t  2 - y t  3 +0.1934 y t  4 Fig. 6.

t 0.2404  t1  0.8924  t  2  0.3196  t  3 2 t  NID(0,0.12633 ). (5)

The simulated wind speed SWt is computed as (6):

SWt t t. y t (6) Effective load carrying capability of Musselroe is investigated in 2014 for the forecast 1936 MW peak load [1]. Applying the same process described before, the capacity value of this 168 MW wind farm is around 38 MW (e.g. 23% of its installed capacity). Although this wind farm has higher installed capacity than the 140MW Woolnorth wind farm, its capacity value is lower. One possible explanation could be the Figure 6. Capacity value of Tasmanian wind farms with and without hydro lower wind speed in Musselroe compared with Woolnorth as power coordination shown in Table I and Fig. 2. As shown in Fig.6, the capacity values of Woolnorth and Musselroe have increased. This means that without using extra water and just by changing the power production plan of the Gordon hydro station, the load carrying capability of wind ACKNOWLEDGMENT farms could be significantly increased. It should be noted that The authors would like to thank M. Willis and G. Carson this increase depends on the wind regime of the wind farms. at Hydro Tasmania for providing hydro storage data and Taking Woolnorth as an example, coordinating 30 MW of Australian Bureau of Meteorology for supplying wind speed hydro power has raised its capacity value from 31% to 46% or data. The authors would like to express their sincere thanks to by 20 MW (from 44 MW to 64 MW). At Musselroe the same Prof. Simon Bartlett for his helpful comments on several coordination has results in an increase from 23% to 33% or by aspects of the paper. 17 MW (from 38 MW to 55 MW). It is considered highly REFERENCES likely that the impact of coordinating on capacity value depends on wind speed regime at each wind farm, with larger [1] “Electricity statement of opportunities for the national electricity market 2011”, Australian Energy Market Operator (AEMO), September increases occurring at wind farms with more consistent wind. 2011.Available: Thus it is concluded that it would be better to use hydro units http://www.aemo.com.au/Electricity/Planning/Reports/~/media/Files/Ot to support wind farms with higher wind speeds if the objective her/planning/2011_Electricity_Statement_of_Opportunities.pdf.ashx is to maximize the benefits from increased capacity value. [2] R. Billinton and D. Huang, “Incorporating wind power in generating capacity reliability evaluation using different models,” IEEE Trans. To investigate the effect of coordination criterion on the Power Systems, vol. 26, pp. 2509–2517, Nov. 2011. capacity value of wind power, the study was repeated with a [3] C. Singh, “Population-based intelligent search in reliability evaluation 50 MW coordination capacity. In this simulation, the hydro of generation systems with wind power penetration,” IEEE Trans. unit could stop power production and retain its water when the Power Systems, vol. 23, pp. 1336–1345, Aug. 2008. [4] R. Billinton and Y. Gao, “Multistate wind energy conversion system output power of the wind farm is more than 50 MW during models for adequacy assessment of generating systems incorporating light load. Although this value is 20 MW higher, the results wind energy,” IEEE Trans. Energy Conversion, vol. 23, pp. 163–170, are similar to the previous study (coordinating 30 MW) and Mar. 2008. have not materially changed. It has increased the capacity [5] A. S. Dobakhshari, S. Member, M. Fotuhi-firuzabad, “A reliability value of Woolnorth and Musselroe wind farms 21 MW and 18 model of large wind farms for power system adequacy studies,” IEEE Trans. Energy Conversion, vol. 24, pp. 792–801, Sep. 2009. MW, respectively. The main reason could be the lower [6] F. Vallee, J. Lobry, and O. Deblecker, “System reliability assessment probability of producing more than 50MW during off peak for method for wind power integration,” IEEE Trans. Power Systems, vol. wind farms as is shown in Table V. 23, pp. 1288–1297, Aug. 2008. [7] B. Hasche, A. Keane, and M. O’Malley, “Capacity value of wind TABLE V. PROBABILITY OF WIND FARMS’ OUTPUT POWER DURING power, calculation, and data requirements: the Irish power system OFF-PEAK case,” IEEE Trans. Power Systems, vol. 26, pp. 420–430, Feb. 2011. Output Power Probability During Off-Peak [8] W. Wangdee and R. Billinton, “Considering load-carrying capability Wind and wind speed correlation of WECS in generation,” IEEE Trans. Farm More than 30 MW More than 50 MW Energy Conversion, vol. 21, pp. 734–741, Sep. 2006. [9] C. Potter and M. Negnevitsky, “Very short-term wind forecasting for Woolnorth 0.68 0.50 Tasmanian power generation,” IEEE Trans. Power Systems, vol. 21, pp. Musselroe 0.51 0.36 965–972, May 2006. [10] R. Karki, P.Hu and R. Billinton, “Reliability evaluation considering wind and hydro power coordination,” IEEE Trans. Power Systems, vol. Therefore, selecting a higher coordination capacity may 25, pp. 685–693, May 2010. not improve the capacity value of wind farms significantly, [11] W. Wangdee, W. Li, and R. Billinton, “Coordinating wind and hydro generation to increase the effective load carrying capability,” in Proc. and an optimum strategy should be selected according to th hydro unit constraints and the output profile of wind farms. 2010 IEEE 11 International Conference on Probabilistic Methods Applied to Power Systems, pp. 337–342. [12] “AEMO Aggregated Price and Demand Data Files.” [Online]. V. CONCLUSIONS Available: http://www.aemo.com.au/Electricity/Data/Price-and- The capacity value of Tasmanian wind farms is calculated Demand. [Accessed: 01-Oct-2012]. [13] “Electricity statement of opportunities for the national electricity in this paper using Monte Carlo simulation technique. The loss market 2012”, Australian Energy Market Operator (AEMO), August of energy expectation is utilized as the reliability index to 2012. Available: compute the capacity value. Furthermore, the impact of http://www.aemo.com.au/Electricity/Planning/Reports/~/media/Files/Ot coordinating a hydro station with these wind farms on the her/planning/2012_Electricity_Statement_of_Opportunities.pdf.ashx effective load carrying capability of wind power is presented. [14] Hydro Tasmania, “Gordon Power Station Technical fact sheets,” 2012. [Online]. Available: http://www.hydro.com.au/energy/our-power- Woolnorth and Musselroe wind farms with different wind stations/gordon-pedder. [Accessed: 05-Oct-2012] regimes are studied and Gordon hydro station is coordinated to [15] “Energy Storage - Historical Data.”, Hydro Tasmania, [Online]. them with various capacities. Wind power is modeled as Available: http://www.hydro.com.au/system/files/water- negative load and ARMA model is used to simulate wind storage/storage_summary-4.xls. [Accessed: 01-Oct-2012]. speed for Musselroe wind farm, which will be added in 2013. [16] A. Keane, et al., “Capacity value of wind power,” IEEE Trans. Power Systems, vol. 26, pp. 564–572, May 2011. It is shown that wind-hydro coordination can improve the [17] R. Billinton and W. Li, “Reliability assessment of electric power capacity value of wind farms regarding their wind regime. systems using Monte Carlo methods”. New York: Plenum Press, 1994, Furthermore, using different capacities to coordinate wind and pp. 22–96. hydro power indicates that coordinating with higher capacities [18] “Vestas Wind Systems.” [Online]. Available: may not result in a great increment in the capacity value of http://www.vestas.com/en/media/brochures.aspx. [Accessed: 01-Oct- wind farms. Therefore, further work should be done to find the 2012]. [19] R. Billinton, H. Chen, and R. Ghajar, “Time-series models for optimum allocated capacity. reliability evaluation of power systems including wind energy,” Microelectronics Reliability, vol. 36, pp. 1253–1261, Sep. 1996.