IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 2, JUNE 2006 533 A Simplified Wind Power Generation Model for Reliability Evaluation Rajesh Karki, Senior Member, IEEE, Po Hu, Student Member, IEEE, and Roy Billinton, Life Fellow, IEEE Abstract—Renewable energy sources, especially wind turbine generators, are considered as important generation alternatives in electric power systems due to their nonexhausted nature and benign environmental effects. The fact that wind power penetration continues to increase has motivated a need to develop more widely applicable methodologies for evaluating the actual benefits of adding wind turbines to conventional generating systems. Reliability evaluation of generating systems with wind energy sources is a complex process. It requires an accurate wind speed forecasting technique for the wind farm site. The method requires historical wind speed data collected over many years for the wind farm location to determine the necessary parameters of the wind speed models for the particular site. The evaluation process should also accurately model the intermittent nature of power output from the wind farm. A sequential Monte Carlo simulation or a multistate wind farm representation approach is often used. This paper presents a simplified method for reliability evaluation of power systems with wind power. The development of a common wind speed model applicable to multiple wind farm locations is presented and illustrated with an example. The method is further simplified by determining the minimum multistate representation for a wind farm generation model in reliability evaluation. The paper presents a six-step common wind speed model applicable to multiple geographic locations and adequate for reliability evaluation of power systems containing significant wind penetration. Case studies on a test system are presented using wind data from Canadian geographic locations. Index Terms—Generation system, reliability evaluation, timeseries model, wind power. I. INTRODUCTION IND is one of the fastest growing energy sources, and is regarded as an important alternative to traditional power generating sources. Enhanced public awareness of the environment has led to rapid wind power growth throughout the world to reduce greenhouse gas emissions associated with conventional energy generation. An energy policy known as the renewable portfolio standard (RPS) is being widely accepted around the world. Acceptance of the RPS is a commitment to produce a specified percentage of the total power generation from renewable sources within a certain date. In the North America, many US states and Canadian provinces have agreed to generate between 5% and 25% of electrical power from renewable energy resources by 2010–2015. Most of this renewable energy will W Manuscript received September 15, 2005; revised September 15, 2005. Paper no. TEC-00177-2005. The authors are with the Power System Research Group, Department of Electrical Engineering, University of Saskatchewan, Saskatoon, SK S7N 5A9 Canada (e-mail: Rajesh.Karki@usask.ca, poh591@mail.usask.ca, Roy Billinton@engr.usask.ca). Digital Object Identifier 10.1109/TEC.2006.874233 come from wind as other renewable sources are not suitable for bulk power generation. The development of comprehensive reliability and costevaluation techniques becomes more important as wind power penetration levels in traditional power systems continue to increase in the near future. Both Monte Carlo simulation [1], [2] and analytical methods [3]–[5] have been utilized in adequacy assessment of generation systems containing wind power. Simulation methods can recognize the chronology of wind variation and its impact on a power system. Reference [6] presents an algorithm to simulate the hourly wind speed using a time-series auto regressive and moving average (ARMA) model. The method requires actual hourly wind speed data collected over a long period of time for the particular geographic location to construct a wind speed simulation model for the specific site. This model can reflect the true probabilistic characteristics of wind speed for the wind site. The process for obtaining the proper ARMA model is quite complex, and is specific to the geographic location of the wind farm. Complex techniques are not readily applied in a practical world. It would be very useful from a practical application point of view if a common ARMA model could be developed that would work for multiple wind farm locations [7]. It should also be noted that historical wind data are not available for all potential wind farm locations. This paper presents the development of a common wind speed model that can generate wind speed probability distributions for multiple wind farm sites if their annual mean wind speed and standard deviation are known. Historical wind speed data from three Canadian sites (Swift Current, North Battleford, and Toronto, which have different geographic characteristics), are used in this paper to obtain the common wind speed model. The wind speed distribution obtained from the common wind speed model can then be used in an analytical method for reliability evaluation of power systems containing wind farms. A wind turbine generator (WTG) is modeled as a multistate unit in reliability evaluation using an analytical method [3]–[5]. The multistate model represents a WTG by a number of derated power output states using the wind speed model for the wind farm location and the power curve of the WTG. The model appropriately reflects the fluctuating characteristics of the sitespecific wind. Reliability studies using the Roy Billinton Test System (RBTS) [8] are presented to determine the appropriate number of states in the multistate WTG model, and develop a simplified model applicable to any wind farm connected to a power system. The development of a simplified 6-step common wind speed model that can be applied to any wind farm in power 0885-8969/$20.00 © 2006 IEEE 534 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 2, JUNE 2006 system reliability evaluation is presented. The validity of the developed method and its practicality is illustrated by evaluating wind power penetration in the RBTS. TABLE I MAIN STATISTICAL CHARACTERISTICS OF SIMULATED AND ACTUAL WIND SPEED DATA II. SITE-SPECIFIC WIND DATA SIMULATION Wind resource at a geographic location is highly variable. Power generated from WTG depends on the wind speed, which fluctuates randomly with time. Wind power studies, therefore, require accurate models to forecast wind speed variation for wind farm locations of interest. Wind speed distributions are often characterized by Weibull distributions, and methods using Weibull distributions [5] have been previously used to estimate wind speed data in wind power studies. Historical hourly data for the wind farm site collected over significant time are normally required to obtain the shaping parameters. An ARMA time-series model has been used by researchers [1], [2], [6] as an accurate method to forecast wind speed data at any particular location. The wind speed at any given time for a particular geographic location can be simulated using (1) (1) SWt = µt + σt . . . yt where SWt is the simulated wind speed for hour t from historical mean wind speed µt and standard deviation σt . Timeseries values of yt can be obtained sequentially using (2), where ϕi(i = 1, . . . , n) and θj(j = 1, . . . , m) are the auto-regressive and moving average parameters of the model, respectively yt = φ1 yt−1 + φ2 yt−2 + . . . + φn yt−n + αt − αt−1 θ1 − αt−2 θ2 − . . . − αt−m θm (2) where {αt } is a normal white noise process with zero mean and a variance of σa2 (i.e., αt ∈ NID(0, σa2 ), where NID denotes normally independently distributed. The order [n, m] and the values of the auto-regressive and moving average parameters shown in (2) can be obtained from many years of collected wind speed data using nonlinear least-square methods [9]. A computer program was developed to estimate the parameters of ARMA (n, m) models for geographic wind farm sites using the algorithm provided in [9]. Historical wind data was obtained for three different locations in Canada: Swift Current, North Battleford, and Toronto. Swift Current lies in the southern part of the Saskatchewan province and is home to one of the largest wind farms in Canada. North Battleford lies to the north in the same province, but does not have a good wind resource. Toronto lies at the great lakes and has a coastal wind regime. These three locations have diverse wind variation patterns and are therefore selected in the study. Historical data on hourly wind speeds for three years (from January 1, 2001 to December 31, 2003), and the hourly mean and standard deviation of wind speeds from a 15-year database (from January 1, 1989 to December 31, 2003) were obtained from Environment Canada to determine the ARMA model for the Swift Current location. The parameters are shown in (3) yt = 0.8782yt−1 − 0.0066yt−2 + 0.0265yt−3 + αt − 0.2162αt−1 + 0.0091αt−2 αt ∈ NID(0, 0.557922 ). (3) Equations (4) and (5) show the time-series ARMA models obtained similarly for North Battleford and Toronto locations respectively yt = 1.7901yt−1 − 0.9087yt−2 + 0.0948yt−3 + αt − 1.0929αt−1 + 0.2892αt−2 (4) αt ∈ NID(0, 0.4747622 ) yt = 0.4709yt−1 + 0.5017yt−2 − 0.0822yt−3 + αt + 0.1876αt−1 − 0.2274αt−2 (5) αt ∈ NID(0, 0.5508 ). 2 Hourly wind speeds were repeatedly simulated for a large number of yearly samples to obtain annual wind speed data for the three sites. Table I shows the mean value and standard deviation of 20 years’ average actual collected data and simulated wind speed data for a year for these sites. It can be seen that the simulated values are very close to those obtained from the actual data. The entire range of the simulated data is equally divided into Na intervals, and the SWai (i = 1, . . . , Na ) is the midpoint of each interval. The probability Pai of a wind speed in an interval SWai (i = 1, . . . , Na ) can be obtained using (6) by counting the number Nai (i = 1, . . . , Na ) of simulated wind speed data in the interval Pai = Nai /(8760 × Ny ) (6) where Ny is the number of sample years simulated. The wind speeds SWai at each of the Na intervals and their corresponding probabilities of occurrence Pai were calculated for the three sites. Fig. 1 shows the probability distributions of simulated wind speed data for these three sites. Wind speed probability distributions are often represented by Weibull distributions. Weibull distribution in general can have various shapes depending on the shaping parameters. The wind speed probability distributions obtained for the three diverse geographic locations, however, are close to normal distribution as shown in Fig. 1. It can however be seen that the three curves are quite different, as the three selected sites have different types of wind regimes. Time-series ARMA model generate both positive and negative values of wind speeds as seen in Fig. 1. Negative values have no physical meaning in this study, and are set to zeros as recommended in [7]. KARKI et al.: A SIMPLIFIED WIND POWER GENERATION MODEL FOR RELIABILITY EVALUATION Fig. 1. Probability distributions of simulated wind data. Fig. 2. III. COMMON WIND SPEED MODEL FOR MULTIPLE WIND SITES The determination of a proper wind speed model for a wind farm location is a complicated process as described in Section II. The process also requires historical wind speed data collected over a significant period of time. It would be very useful for practical applications, if a common wind speed model could be used for benefit analysis of wind power sources located at different geographic sites. This approach will prove useful for prospective wind farm locations lacking adequate historical data. The objective is to determine a general model that can be used to obtain power output from wind turbines located at different sites with reasonable accuracy. The probability distributions of wind speeds at the three selected geographic sites are shown in Fig. 1 in Section II. In this section, the probability distributions are plotted in terms of the annual mean wind speed value µ, and the standard deviation σ for the selected locations. The distribution considers wind speeds up to 10 σ to include extreme values despite their low probability of occurrences. The distribution is divided into Nb number of interval steps, and each step has a length of 10 σ/Nb , where the midpoint values of these steps are SWbi (i = 1, . . . , Nb ), and one of the steps has a midpoint value of µ. The steps SWbi are defined as in (7). For example, for a 100-step model, the midpoint values SWbi of the wind speed steps will be as follows: µ–4.9σ, µ–4.8σ, . . . , µ–0.1σ, µ, µ + 0.1σ, µ + 0.2σ, . . . , µ + 5σ. The probability of each step Pbi (i = 1, . . . , Nb ) can be calculated by (8) SWbi = µ + (10σ/Nb ) × (i − 0.5 × Nb ) for even Nb , and = µ + (10σ/Nb ) × (i − 0.5 × (Nb + 1)) for odd Nb (7) Pbi = Nbi /(8760 × Ny ) (8) where Nbi is the number of simulated wind speed data in the step SWbi (= 1 . . . , Nb ). 535 Combining wind speed models for different sites. The probability distributions plotted in terms of µ and σ as discussed above are shown in Fig. 2 for the three locations. It can be seen that the three curves are very close. The three curves for the three different wind regimes can be combined to obtain a common wind speed model. The probability Pci (i = 1, . . . , Nb ) for each step in the common wind speed model is obtained by taking the average value of Pbi (i = 1, . . . , Nb ) for these three wind sites. Fig. 2 also shows the curve for the common wind speed model. The common wind speed model can be used to obtain the wind speed probability distribution for any geographic location with wind speed characteristics close to the selected Canadian sites. The only data required are the annual mean wind speed µ and the standard deviation σ for that site. Negative wind speed values are converted to zeroes in this process. It is shown in Section VI that the wind speeds and their probabilities obtained for a specific site using the common wind speed model can be used with reasonable accuracy in the reliability evaluation of power systems containing wind power. Fig. 3 shows a common wind speed model that can be used for multiple geographic locations. The application of the common wind speed model in Fig. 3 is illustrated in this paper by applying it to a site in Regina, Saskatchewan. The µ and σ values for Regina wind data are 19.53 and 10.06 km/h, respectively. The common wind speed model is used to obtain the wind speeds and their corresponding probabilities for the example site. Table II shows the wind speeds in km/h and their corresponding probabilities. Only 5 out of the 100 steps in the wind speed model are shown. It should be noted that the negative values are converted to zeroes in Table II. IV. COMMON WTG POWER GENERATION MODEL FOR RELIABILITY EVALUATION It is important to accurately assess the electric power generated by a WTG located at a particular geographic site in reliability analysis. The generated power varies with the wind 536 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 2, JUNE 2006 Fig. 5. Fig. 3. Common wind speed model. TABLE II WIND MODEL FOR REGINA SASKATCHEWAN Development of a WTG model. Fig. 5 shows the development of the power generation model of a WTG located at a particular geographic site. The annual µ and σ wind speed data for the wind farm site is used in the common wind speed model shown in Fig. 3 to obtain the sitespecific wind speed model. This site-specific wind speed model is then combined with the WTG power curve to obtain the WTG power generation model. Equation (9) is the mathematical expression for the power curve. The power generated Pi (i = 1, . . . , Nb ) corresponding to a given wind speed SWbi (i = 1, . . . , Nb ) can therefore be obtained from (9) 0 ≤ SWbi < Vci Pi = 0, = Pr (A + B × SWbi + C × SW2bi ), Vci ≤ SWbi < Vr Vr ≤ SWbi ≤ Vco = Pr , = 0, Vco < SWbi (9) where the constants A, B, and C are presented in [3]. Wind speeds less than Vci and greater than Vco produce zero output power from the WTG. The wind speed model steps corresponding to these values can, therefore, be combined into a cumulative step, and the zero output probability Pp0 can be obtained from (10). Similarly, wind speed steps between Vr and Vco can be combined, and the probability of rated power output Ppr can be obtained from (11) Pci for 0 ≤ SWbi < Vci or Vco < SWbi Pp0 = Ppr = Fig. 4. (10) Pci for Vr ≤ SWbi ≤ Vco (11) Power curve of a WTG. speed at the wind farm site. The power output of a WTG can be determined from its power curve, which is a plot of output power against wind speed. Fig. 4 shows a typical power curve of a WTG. A WTG is designed to start generating at the cut-in speed Vci (Vc ) and is shut down for safety reasons at the cut-out speed Vco . Rated power Pr is generated when the wind speed is between the rated speed Vr and the cut-out speed Vco . There is a nonlinear relationship between the power output and the wind speed when the wind speed lies within the cut-in speed Vci and the rated speed Vr as shown in Fig. 4. The application of the common WTG power generation model is illustrated in this paper by applying it to a WTG installed in Regina. WTG rated at 1.5 MW, and with cut-in, rated, and cut-out wind speeds of 14.4, 45, and 90 km/h, respectively, are used in the studies. The site-specific wind speed model shown in Table II is combined with the power curve to obtain the WTG generation model. Table III shows the power output levels and their corresponding probabilities. The 100step wind speed model in Table II is reduced to 33 steps in the power generation model in Table III as different wind speeds producing the same power level are aggregated into single steps. KARKI et al.: A SIMPLIFIED WIND POWER GENERATION MODEL FOR RELIABILITY EVALUATION TABLE III POWER GENERATION MODEL FOR A WTG IN REGINA 537 TABLE V LOLE COMPARISON OF RBTS INCLUDING WECS LOCATED AT DIFFERENT WIND SITES TABLE IV TRADITIONAL GENERATING UNIT RELIABILITY DATA (RBTS) A wind farm usually consists of many WTG units subjected to the same wind regime characterized by the geographic location. The total power generation model for the wind farm can be obtained by summing up the individual WTG power levels corresponding to each wind speed interval in the common wind speed model. V. VALIDITY OF THE COMMON WIND SPEED MODEL FOR RELIABILITY STUDIES The power generation model for a wind farm can be easily obtained from the common wind speed model shown in Fig. 3 using only the site-specific µ and σ data and the WTG power curve. The validity of this common model for practical reliability applications is analyzed using a test system. Reliability studies using analytical techniques were conducted on the RBTS [8] to illustrate the practicality of a common wind speed model for different sites. The RBTS is a basic reliability test system that evolved from the research and teaching program conducted at the University of Saskatchewan. The RBTS consists of 11 traditional generating units with a total capacity of 240 MW. The generating unit ratings and reliability data are shown in Table IV. The system peak load is 185 MW [8]. A wind farm with 18 WTG units was added to the RBTS. WTG rated at 1.5 MW, and with cut-in, rated, and cut-out wind speeds of 14.4, 45, and 90 km/h, respectively, was used in the studies. The added wind capacity is 27 MW, which is almost 10% of the total installed system capacity. The generating system adequacy of the test system was evaluated for three different cases, in which the wind farm was considered at separate locations with North Battleford, Toronto, and Swift Current wind regimes. Table I shows the mean wind speed and the standard deviation for the three sites. Table V shows the loss of load expectation (LOLE) [10] for the RBTS before and after adding WTG at the different locations. The LOLE results were obtained by convolving the Fig. 6. LOLE comparison for North Battleford using its ARMA model and the common wind speed model. system generation model with the load model using an analytical technique. The conventional generating units were represented by a two-state model, in which the unit is either operating at rated power output or on forced outage. To account for the randomly varying power output from WTG units, they were represented by 100-state model in the studies. The annual load duration curve for the RBTS was used as the load model, and the data are provided in [8]. The second row of Table V shows the indices using the site-specific ARMA models shown in (3)–(5) for the three locations. The third row shows the results obtained using the common wind speed model in Fig. 3. It can be seen that the results are very close. It can also be seen from Table V that the adequacy of the RBTS is improved by adding the 27MW wind farm. The improvement with the Swift Current site is more significant than that of the other two locations, as Swift Current has a higher mean wind speed, and therefore, provides a better wind resource. Table V shows that the common wind speed model produces accurate results for wind farm locations with different wind regimes. Power system reliability is very sensitive to changes in the system peak load. The accuracy of the common wind speed model is therefore analyzed for different peak load levels at the three geographic wind farm locations. Fig. 6 shows the system LOLE for different peak load levels with the wind farm located in North Battleford. The results obtained using the common wind speed model is compared with those obtained using the North Battleford ARMA model given in (4). Fig. 6 shows that the common wind speed model also produces very close results at different peak load levels. Fig. 7 shows similar results when the wind farm is considered to be in the Toronto location. The results show that the common wind speed model can be utilized in the reliability evaluation of power systems containing wind farms, 538 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 2, JUNE 2006 Fig. 7. LOLE comparison for the Toronto Data using its ARMA model and the common wind speed model. Fig. 8. Swift Current wind site with a 10% wind power penetration level (18 WTG). and provide very close results compared to the site-specific ARMA models for the different sites. VI. SIMPLIFIED MULTISTATE WTG MODEL A conventional generating unit is usually represented by a two-state model in power system reliability evaluation. The two states are the unit-operating state with rated power output, and the unit-failure state with zero output. A WTG unit, however, cannot be represented by a two-state model since the power output can vary continuously and intermittently from zero to the rated value depending on the wind speed at the wind farm site. WTG units are therefore represented by multistate models in analytical methods used in power system reliability assessment. Accurate reliability assessment can be obtained by representing a wind farm by a large number of discrete output states. This can be obtained from a wind speed model with a large number of discrete wind speed steps. The model can be simplified by reducing the number of steps at the cost of accuracy. It is, therefore, important to determine the minimum number of steps that can be used in a simplified wind speed model to obtain reasonable accuracy for all practical purposes. Studies were done to evaluate the system LOLE with increasing peak load levels using different numbers of wind speed steps in the common wind speed model. The wind farm is assumed to be located in Swift Current. Fig. 8 shows the results of the study. It can be seen that using six or more steps in the wind speed model provides close results. The reliability contribution of a wind farm highly depends on the site wind regime. Studies similar to that shown in Fig. 8 were also conducted for a wind farm site with a different wind regime. Fig. 9 shows the system LOLE results for the case in which the wind farm is assumed to be located in Toronto. Fig. 9 also shows that a 6-step wind speed model is adequate for reliability studies with reasonable accuracy. It should be noted that 27 MW of wind capacity is added to the RBTS in the above studies, which amounts to about a 10% wind penetration. Since the reliability contribution of wind sources is highly dependent on the wind penetration level, it is important to determine the appropriate number of wind speed model steps at practical wind penetration levels. The existing wind penetration levels in most power systems are much lower than 10% for Fig. 9. Toronto wind site with a 10% wind power penetration level (18 WTG). Fig. 10. Swift Current wind site with a 15% wind power penetration level (28 WTG). which the above results are obtained. Studies conducted for wind penetration of 1% showed that the number of steps had no impact on the system LOLE results. When the penetration level is insignificant, wind power has relatively little influence on the total system reliability performance. Reliability evaluation in power system planning should consider a long-term planning horizon. It is expected that wind power penetration will grow to 10%–15% in the next 10–15 years. Fig. 10 shows the results obtained when the wind farm added to the RBTS contains 28 WTG units, increasing the wind penetration to 15% of the total installed system capacity. The results show that a six-step wind speed model is adequate for reliability studies with reasonable accuracy. KARKI et al.: A SIMPLIFIED WIND POWER GENERATION MODEL FOR RELIABILITY EVALUATION Fig. 11. Six-step common wind speed model. Fig. 12. index. 539 Comparison of different models for WTG power output by LOLE TABLE VI Six-STEP WIND MODEL FOR REGINA SITE using the simplified model in Table VII. The results obtained from the simplified model are compared with those obtained from a 100-step wind speed model derived from an accurate ARMA (4, 3) model for Regina as shown in (13) yt = 0.9336yt−1 + 0.4506yt−2 − 0.5545yt−3 + 0.1110yt−4 + αt − 0.2033αt−1 − 0.4684αt−2 + 0.2301αt−3 (13) αt ∈ NID(0, 0.409423 ) 2 TABLE VII SIMPLIFIED POWER GENERATION MODEL FOR THE REGINA WIND FARM The 6-step common wind speed model that can be used for reliability studies of power systems with reasonable accuracy is shown in Fig. 11. The model shows the six wind speed steps and the corresponding probabilities. The wind speeds for a geographic site can be obtained from this model using (12), if µ and σ for the site are known SWbi = µ + (i − 3) × (5σ/3) for (i = 1, . . . , 6). (12) The application of the six-step common wind speed model is illustrated by applying it to a site in Regina. Table VI shows the wind speeds in km/h and their corresponding probabilities. It should be noted that the negative values are converted to zeroes. A simplified multistate power generation model for a WTG can be determined by combining the 6-step common wind speed model with the power curve equation (9). This method can be utilized to develop a simple but adequate multistate wind generation model. The only data required are the wind farm site annual µ and σ data, and the power-curve data for the WTG units. For example, Table VII shows the simplified multistate power generation model for a wind farm consisting of 18 WTG units in Regina. Fig. 12 shows the results of reliability assessment of the RBTS system with 10% wind penetration coming from the 27-MW Regina wind farm. The shaded bars show the LOLE results It can be seen from Fig. 12 that the simplified multistate WTG power generation model derived from the common 6-step wind speed model can provide very close results when compared with the 100-step model derived from the site-specific time-series ARMA wind speed model for Regina. The simplified model can therefore be used in the reliability evaluation of generating system including wind power with reasonable accuracy. VII. CONCLUSION It becomes increasingly important to develop realistic reliability evaluation techniques that are practically useful for electric power industries that are expected to include a rapidly growing proportion of wind generation in the coming years. The benefits from wind sources are largely dictated by the wind regime at the wind farm site. It is, therefore, very important to obtain suitable wind speed simulation models and appropriate techniques to develop power generation model for WTG in reliability evaluation. It requires a significant amount of historical data and effort to develop a realistic wind speed model for a geographic site. Historical data from three Canadian sites with diverse wind regimes were used to obtain a common wind speed model in this paper. The common wind speed model can be applied to obtain a wind farm power generation model for any geographic location if the mean wind speed and standard deviation, and the WTG power curve parameters are known. Results from different analytical reliability studies are presented in this paper to show that a wind power generation model can be greatly simplified by using only 6 steps in the common wind speed model. The technique to develop the simplified WTG multistate model for any geographic location is presented and illustrated with an example. The results show that the simplified method 540 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 21, NO. 2, JUNE 2006 can be used in the reliability evaluation of a generating system including wind power with reasonable accuracy. This approach will be very useful for wind farm locations lacking adequate historical data. The presented simplified method can be easily used for practical applications. REFERENCES [1] R. Karki and R. Billinton, “Cost-effective wind energy utilization for reliable power supply,” IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 435–440, Jun. 2004. [2] R. Billinton and G. Bai, “Generating capacity adequacy associated with wind energy,” IEEE Trans. Energy Convers., vol. 19, no. 3, pp. 641–646, Sep. 2004. [3] P. Giorsetto and K. F. Utsurogi, “Development of a new procedure for reliability modeling of wind turbine generators,” IEEE Trans. PAS, vol. 102, no. 1, pp. 134–143, Jan. 1983. [4] X. Wang, H. Dai, and R. J. Thomas, “Reliability modeling of large wind farms and electric utility interface systems,” IEEE Trans. PAS, vol. 103, no. 3, pp. 569–575, Mar. 1984. [5] I. Abouzahr and R. Ramakumar, “An approach to assess the performance of utility-interactive wind electric conversion systems,” IEEE Trans. Energy Convers., vol. 6, no. 4, pp. 627–638, Dec. 1991. [6] R. Billinton, H. Chen, and R. Ghajar, “Time-series models for reliability evaluation of power systems including wind energy,” Microelectron. Reliab., vol. 36, no. 9, pp. 1253–1261, 1996. [7] R. Karki and P. Hu, “Wind power simulation model for reliability evaluation,” in Proc. IEEE Can. Conf. Electr. Comput. Eng., Saskatoon, May 1–4, 2005, pp. 541–544. [8] R. Billinton and S. Kumar, “A reliability test system for educational purposes—Basic data,” IEEE Trans. Power Syst., vol. 4, no. 3, pp. 1238– 1244, Aug. 1989. [9] S. M. Pandit and S. M. Wu, Time Series and System Analysis With Application. New York: Wiley, 1983. [10] R. Billinton and R. N. Allan, Reliability Evaluation of Power Systems, New York: Plenum, 1996. Rajesh Karki (S’98–M’00–SM’04) received the B.E. degree from Burdwan University, Durgapur, India, in 1991, and the M.Sc. and Ph.D. degrees from the University of Saskatchewan, Saskatoon, SK, Canada in 1997 and 2000, respectively. He was a Lecturer at the Tribhuvan University, Kathmandu, Nepal. He was also an Electrical Engineer with Nepal Hydro & Electric, Butwal, Nepal; Udayapur Cement Industries, Udayapur, Nepal; Nepal Telecommunications Corporation, Kathmandu, Nepal; and General Electric Canada, Peterborough, ON, Canada. Currently, he is an Assistant Professor in the Department of Electrical Engineering at the University of Saskatchewan. Po Hu (S’04) received the B.E. degree from the Chongqing University, Chongqing, China, in 2001 and the M.Sc. degree from the University of Saskatchewan, Saskatoon, SK, Canada in (2005). He is currently pursuing the Ph.D. degree at the same University. Roy Billinton (S’59–M’64–SM’73–M’78–LF’01) received the B.Sc. and M.Sc. degrees from the University of Manitoba, Winnipeg, MB, Canada, in 1960 and 1963, respectively and the Ph.D. and D.Sc. degrees from the University of Saskatchewan, Saskatoon, SK, Canada in 1967 and 1975, respectively. In 1964, he joined the University of Saskatchewan. He was also with Manitoba Hydro, Winnipeg, MB, Canada, in the System Planning and Production Divisions. He is Formerly Acting Dean of Graduate Studies, Research and Extension of the College of Engineering at the University of Saskatchewan. Currently, he is a Professor Emeritus in the Department of Electrical Engineering. He is also an author of power system analysis, stability, economic system operation, and reliability papers. Dr. Billinton is a Fellow of the EIC and the Royal Society of Canada.