Stochastic Performance Assessment and Sizing for a
Hybrid Power System of Solar/Wind/Energy Storage
Amirsaman Arabali, Student Member, IEEE, Mahmoud Ghofrani, Member, IEEE, Mehdi Etezadi-Amoli, Life Senior Member, IEEE, and Mohammed Sami Fadali, Senior Member, IEEE AbstractThis paper proposes a stochastic framework for optimal sizing and reliability analysis of a hybrid power system including the renewable resources and energy storage system. Uncertainties of wind power, photovoltaic (PV) power, and load are stochastically modeled using autoregressive moving average (ARMA). A pattern search-based optimization method is used in conjunction with a sequential Monte Carlo simulation (SMCS) to minimize the system cost and satisfy the reliability requirements. The SMCS simulates the chronological behavior of the system and calculates the reliability indices from a series of simulated experi- ments. Load shifting strategies are proposed to provide some exibility and reduce the mismatch between the renewable genera- tion and heating ventilation and air conditioning loads in a hybrid power system. Different percentages of load shifting and their potential impacts on the hybrid power system reliability/cost anal- ysis are evaluated. Using a compromise-solution method, the best compromise between the reliability and cost is realized for the hybrid power system. Index TermsLoad shifting strategy, sequential Monte Carlo simulation (SMCS), stochastic modeling. NOMENCLATURE , Order of the autoregressive moving average (ARMA) model. Area for the photovoltaic (PV) array. th decision criterion. Capital cost of energy/power for storage system. Cost of hybrid power system. Self-discharge rate for the energy storage. EE Expected energy demand. ENS Energy not supplied. EENS Expected energy not supplied. EIR Energy index of reliability. PDF for the th simulated random variable. Net power generated by PV. Net power generated by wind. Electric power generated by PV for time . Electric power generated by wind for time . Rated wind power. Power generated by PV. Power generated by wind. Extra-terrestrial radiation on a horizontal surface. Installation cost for PV/wind. Load/modied load at time . Interrupted/shifted load at time . Cost of operation and maintenance for PV. Cost of operation and maintenance for wind. Parameter in the functional form for the . Power rating for the energy storage. Power of the energy storage at time . Parameter in the functional form for the . Ratio, diffuse radiation in hour/diffuse in day. Geometric factor, the ratio of beam radiation on the tilted surface to that on a horizontal surface. Storage system capacity. SW, Sk, SL Simulated wind speed, clearness index, load. Stored energy at time . Wind speed. , and Cut-in, cut-out, and rated wind speed. Sample for the th simulation and th scheduling period. th stochastic input variable. PV/wind capacity installation. Percentage of load shifting. Powerelectronic interface efciency for PV/wind generation. Storage system round-trip efciency. PV array efciency. Moving average/autoregressive coefcient of the ARMA model. , Mean and standard deviation for clearness index. , Mean and standard deviation for the load. , Mean and standard deviation for the wind speed. Satisfaction level for the th decision criterion. Reference level of achievement for the th decision criterion. Reectance of the ground. Set of tradeoff points. I. INTRODUCTION R ENEWABLE resources in the United States have become an important portion of generation mix with a capacity of 12.7% of net electricity generation at the end of 2011. State initiatives, including federal production tax credit and renewable portfolio standards, and the climate change concerns have resulted in the growth of renewable generation capacity of approximately 39% of the national electric capacity additions Manuscript received March 12, 2013; revised June 21, 2013, August 27, 2013, and October 11, 2013; accepted October 23, 2013. Date of publication December 18, 2013; date of current version March 18, 2014. The authors are with the University of Nevada, Reno, NV 89557 USA, and also with the University of Washington, Bothell, WA 98011 USA (e-mail: aarabali@unr.edu; mghofrani@uwb.edu; etezadi@unr.edu; fadali@unr.edu). Digital Object Identier 10.1109/TSTE.2013.2288083 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 2, APRIL 2014 363 1949-3029 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. in 2011 [1]. The stochastic nature of intermittent renewable energy resources complicates their integration of operation and planning purposes. A hybrid power system that is composed of energy storage and renewable generation can alleviate the issues associated with the renewable power supply uctuations. Storage of the renewable power generation in excess of the load makes it accessible for later release when the renewable genera- tion is insufcient to supply the load [2]. This offsets the renewable energyvariations and makes the renewable generation integration more reliable. This is particularly true for exible heating ventilation and air conditioning (HVAC) loads which can be deferred and adopted based on the renewable energy uctuations. However, a large number of random variables and parameters in a hybrid power system makes its optimal design more difcult [3]. This optimization includes optimal compo- nent-sizing for the hybrid system as well as the assessment of system performance to realize the best compromise between the cost and reliability. The reliability study is based on stochastic models to consider the uncertainties of the hybrid power system such as equipment failures and random variations in both generation and load [4]. The design, simulation, and optimization of the isolated hybrid power systems have beenstudiedin[5][8]. References [5] and[6] utilized practical and experimental estimates for component- sizing and performance evaluation without optimizing their parameter values. Optimization techniques are employed in [7] and [8] to design and operate the stand-alone hybrid power systems. However, none of these studies evaluated the reliability of the hybridpower systems for the grid-independent applications. Several studies have proposed probabilistic frameworks for opti- mal sizingand reliabilityanalysis of hybridpower systems [3], [4], [9][14]. References [3], [4], [9], and [10] employed analytical methods to mathematically model the hybrid power system and used direct solutions to evaluate the reliability indices from its model. However, increasing the number of random parameters in a hybrid power systemcomplicates the mathematical computa- tions of the reliability indices. In addition, the analytical methods do not properly represent the random nature of hybrid power systems and their component failures [11]. Monte Carlo simula- tion (MCS) randomly samples the system states, irrespective of event chronologies, to provide a detailed knowledge of the probability distributions of reliability indices. MCS has been used to evaluate the hybrid power system reliability [11][14]. The utilities may have some agreements with the consumers to shift part of their HVACloads during peak load hours. This agreement will provide the customer with economic incentives and prevent the need for additional conventional generation during these peak hours. For example, NVEnergys cool share programis associated with approximately 145 MW remote controlled air conditioning load. These loads can be dispatched by NVEnergy through raise/ lower thermostat commands as needed [15]. A smart strategy is required to implement the load shifting and reduce the mismatch between the renewable generation and HVAC loads. None of these studies investigated load shifting and its potential impact on the hybrid system reliability/cost analysis. This paper proposes a stochastic optimization method to optimally size the hybrid system components while satisfying the system reliability requirements. A sequential Monte Carlo simulation (SMCS) method is used to generate data in a time sequence and then calculate reliability indices fromthe simulated experiments. The SMCS approach simulates the chronological performance of the system by generating sequential samples of system states for numerous time intervals. This provides a better representation of the hybrid systems behavior for reliability analysis when compared to the analytical methods or nonsequen- tial MCS used in the earlier work [3], [4], [9][14]. A direct search tool, pattern search (PS), is utilized to solve the non- smooth problem of optimal sizing for the hybrid power system. The traditional derivative-based optimization methods cannot be used here due to the terms such as max and absolute value that are utilized to calculate a dependent variable in the objective function. In order to improve the computational efciency and speed up the search process, the initial solution is generated using genetic algorithm (GA). A sensitivity analysis based on the epsilon-constraint approach and compromise-solution method is used to realize the best tradeoff between the cost and reliability for the hybrid power system. Our optimization gives the eco- nomical combination of PV and wind generation for satisfying reliability and other physical constraints. Section II explains the PS optimization and SMCS methodol- ogy and the physical or operational constraints on the hybrid power system parameters. Also, ARMA models of PV and wind power, HVAC load as well as our sensitivity analysis are pre- sented in Section II. Different scenarios and their simulation results are giveninSectionIII. SectionIVpresents the conclusion. II. METHODOLOGY The main objective of this paper is to match the HVAC loads with the renewable energy sources without the need for supple- mentary conventional generations or grid-connection strategies. This will expand the use of renewable energy to reduce green- house gases associated with the conventional generation. An isolated stand-alone hybrid energy system including storage, wind, and PV generation is designed to economically and reliably supply the HVAC load. Load shifting is used to provide the demand exibility and meet the objective. A. ARMA Model for Wind Speed, Solar Irradiance, and Load The SMCS requires sampling from sequences of distributions to chronologically simulate the random behavior of the system [16]. Time series methods are used to provide such sequences and model the uncertainties for the sequential analysis. Several refer- ences used ARMA as a time series model to characterize random variations of the wind and solar generation. An th-order auto- regressive (AR) and th-order moving average (MA) model represents the long-termwind/solar/load characteristics [17], [18] where is the white noise with zero mean, zero autocorrelation, and unity variance. Ten years of historical hourly wind speed, clearness index, and load data are used to calculate the AR and MA coefcients and obtain a statistical model that accounts for a long-term scheduling period. This model uses the randomly generated value of and the previous values of and to provide 364 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 2, APRIL 2014 a time series representation for and generate hourly samples for the 1-year scheduling period. Since wind, PV, and load follow the specied patterns and the utilized 10-year data are more likely to cover all the states that may occur for wind, PV, and load proles, the modeling technique properly represents the systems behavior for the scheduling period. The time series ( ) is shifted using the mean and scaled using the standard deviation of the historical data to obtain hourly data for wind speed, clearness index, and load [19] The simulated wind speed is used to calculate the wind power output [20] Assuming an efciency of for the powerelectronic inter- face, the net wind power output is A single-row layout with sufcient distance between the individual turbines is assumed for the wind farm to avoid the wake effects in the prevailing wind direction. The solar irradiance is stochastically modeled using the clearness index ( ) which is described as the quotient of the irradiance on a at level ( ) and the extraterrestrial solar irradiance ( ) [21] The irradiance with an inclining angle of with respect to the horizontal plane is given by Here and are given by where and are the parameters dening the correlation of the diffuse fraction ( ) with the clearness index ( ) as follows [3]: The simulated clearness index is used to calculate the PV power output by Assuming an efciency of for the powerelectronic interface, the net PV power output is The efciency curves of the converters are used to calculate and [22], [23]. B. Optimization Problem The wind and PV capacity installations and are multiplied by the normalized renewable generation to represent the electric power generated by renewable resources. These parameters as well as the power rating and capacity of the storage systemare included in the objective function to minimize the sum of installation, operation, and maintenance costs of the hybrid energy system. The optimization problem can be formulated as follows: where , , , , and are the objective function, equality constraint, inequality constraint, dependent, and decision variables, respectively. The objective function of (14) is given by The last two terms of ( ) indicate the storage cost. The balance-of-system costs for energy and power are included in and , respectively [24]. The powerelectronic interface cost for connecting the PV and wind to the network is considered as a part of and . The constraints of (15) and (16) include design, load, and reliability constraints which are discussed below. C. Design Constraints Because of the physical limits existing in the operation of hybrid system, any feasible solution of the optimization problem must satisfy the following constraints: 1) Storage Constraints: These constraints must satisfy an instantaneous energy balance between the sum of the generated and stored energy and the load. The balance of energy for the charging and discharging of the energy storage system is given by (18a) and (18b), respectively ARABALI et al.: STOCHASTIC PERFORMANCE ASSESSMENT AND SIZING FOR A HYBRID POWER SYSTEM OF SOLAR/WIND/ENERGY STORAGE 365 The storage at time satises the minimum and maximum capacity requirements as The storage system capacity ( ) is dened as the maximum of the storage capacities over the scheduling period Assuming as the power rating, then for each hour, the storage system must satisfy The storage power rating is 2) Load Constraints: Limited human control on the generation of the intermittent renewable resources may lead to a signicant mismatch between the renewable generation and HVACload in a hybrid power system. Load shifting and interruption provide some exibility to reduce this mismatch, especially during periods of insufcient generation and/or peak load. This is particularly true for HVAC loads such as refrigerators, pool pumps, and hot tubes, which have a exible time of operation. Load shifting and interruption are implemented as The shifted load is supplied at the earliest opportunity when the renewable generation is in excess of the required load. The following inequality constraint ensures that the shifted load remains between the minimum and maximum allowable load shifting for time : If , then (18a) is used to charge the storage system. If < , then the storage system is discharged using (18b) to supply the excess load. If the stored energy is less than , then part of the load is shifted to the next hour using (23a) and (23b). If the maximum allowable load shifting ( ) cannot satisfy the minimum capacity requirement specied by (19), then the load is interrupted to balance the hybrid power and load. 3) Reliability Constraints: The reliability assessment evaluates the quality of meeting load in a composite power system. During periods of generation deciency or peak load when the hybrid power system cannot supply the load, load shifting is applied to balance the hybrid power and load. The load is interrupted if the maximum allowable load shifting cannot meet this objective. The reliability assessment of the hybrid power system considers the system uncertainties contributing to the load interruption. These uncertainties include equipment failure rates and random changes in both the renewable generation and load. Equipment failure and repair characteristics provide the two-state models for each individual generating unit. The reliability models of wind turbines and PV arrays in a hybrid power system are the combination of these two-state models and power output models dened by (6) and (13). The wind/PV power output from each individual wind turbine/PV panel is then added to calculate the total power generated by the wind turbines/PVunits within a wind/PV farm. Fig. 1 shows the reliability model for a wind/PV generator. The equipment unavailability or equipment forced outage rate (FOR) is the statistically calculated forced outage probability of the unit for a long time period where is the down time and is the up time of each unit. Both are exponentially distributed. An updown sequence is generat- ed for the scheduling period. This sequence is used with the wind speed/clearness indexpower curves to obtain the generating capacity sequence [25]. The energy not supplied (ENS) over a scheduling period ( ) is The expected energy not supplied (EENS) is a probabilistic index for the reliability analysis of the hybrid power system and is expressed as The reliability of the hybrid power system is measured using the energy index of reliability (EIR) as follows: where EE is the expected energy demand of the hybrid power system design candidates over the scheduling period. A reliability constraint guarantees that the EIR (or EENS) satises the desired reliability level for the hybrid power system. This constraint can be expressed by either of the following inequalities: D. Sequential Monte Carlo Simulation An SMCS is used in this paper to deal with the uncertainties of the hybrid power system. The method simulates the chrono- logical progress of the system by generating the random Fig. 1. Reliability model for wind/PV generator. 366 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 2, APRIL 2014 sequential samples of system states. The ARMA models of renewable energy generation and HVAC loads are used to simulate these sequences. This process is repeated for many trials to obtain distributions for the output randomvariables [25], [26]. The output and input random vectors are related by An SMCS produces a sample ( ) for each input random variable ( ) using its ARMA model. The functional relation- ship between and is then used to produce the output for each sample. It repeats this process for a specied number of simulation ( ) over a scheduling period ( ). Using the sample outputs for the th input random variable ( ), the distribution for the th output random variable ( ) can be derived [27]. Mathematically, an SMCS includes a set of algorithms that produce a collection of weighted random samples , where > and at each time instance such that for the function E. Pattern Search A PS-based optimization technique is utilized in this paper to optimally size wind generation, PV generation, and adequately size energy storage capacity in a hybrid power system, while satisfying a specied reliability level. Generally, an initial start- ing point supplied by the user is used to start the algorithm. A mesh, a set of points, is then established around the present point of the algorithm. The present point is added with a set of vectors, called the direction vectors, which are scaled with a scalar coefcient to make the mesh. The scalar is called the mesh size and its default value is unity for the rst iteration. This is illustrated in Fig. 2 where [0 1], [1 0], , and are the direction vectors and is the initial point. The mesh points are polled by the algorithm based on their tness function values. If a point improves the tness function value over the present point, it is chosen as the present point ( ) for the next iteration and the poll is successful. Upon a successful poll, the algorithm expands the existing mesh size with an expansion coefcient (default value of 2) and uses the new mesh size to form the expanded mesh around the present point. This results in , , , and as the mesh points for iteration 2. A successful poll of the mesh points by the algorithm gives as the present point and the current mesh size is multiplied by 2 to get a mesh size of 4 at the third iteration. Otherwise, if the poll is unsuccessful for iteration 2, the present point is retained ( ) and the mesh size is reduced for the next iteration. Toward this end, the algorithm multiplies the current mesh size by a contraction factor (0.5) and polls the next iteration with a smaller mesh size. The process halts upon the satisfaction of a termination criterion with one of the following situations. 1) The mesh size is smaller than a specied value for the mesh tolerance. 2) The iteration number is greater than a specied value. 3) The distance between the points found at two consecutive successful polls is less than a specied tolerance. This evolutionary algorithm is appropriate to solve the non- linear, nonconvex, and nonsmooth problemof the optimal hybrid power system design, where the classical optimization methods cannot handle the nonconvexity and discontinuity of the objec- tive function. Moreover, the uncertain factors considered for the reliability assessment of the hybrid power system make our optimization problema stochastic programthat cannot be solved by the traditional derivative-based optimization methods. A GA is used to generate the initial solution because of its global search ability which ensures the convergence of the optimization to the global optimum. The direct search tools such as PS are faster than GA to locally nd the optimal solution. The search capability of the population-based evolutionary algo- rithms such as GA is highly dependent on the population size which signicantly increases the computational burden if used in the combination of SMCS for the whole optimization process. Thus, the GA-based SMCS optimization necessitates a compro- mise between the number of simulations and population size to make its computational time reasonable. This compromise may result in an in-convergence state or nonoptimal solution. The local search ability and high speed of PS make it an efcient direct search tool to resume the optimization process from the initial solution generated by GA and improve the computational efciency [28], [29]. We rst ran the optimization problem using our proposed method (PS-based SMCS) and then solved the same problem using GA with the population size of 50 and generation number of 20. Table VI of the Appendix shows the simulation results for the PS-based optimization and the GA results for 1000 simulations. These results demonstrate the improved efciency of the proposed PS-based SMCS over the GA-based SMCS optimization. 1) PS-Based SMCS Optimization: Fig. 3 presents the owchart for the developed method. The vectors of input and output random variables are Ten years of hourly load, clearness index, and wind speed data are used to characterize the uncertainties of the HVAC load and solar/wind generation by their ARMA models. The scaling parameters ( and ) are initialized using GA and then used as the initial solutions to start the PS algorithm. For each sample, a large penalty coefcient ( ) is enforced on the violated constraints to assign a higher cost to an infeasible solution [30]. The combination is given by Fig. 2. Formation of mesh for PS. ARABALI et al.: STOCHASTIC PERFORMANCE ASSESSMENT AND SIZING FOR A HYBRID POWER SYSTEM OF SOLAR/WIND/ENERGY STORAGE 367 The PS optimization obtains the optimal sizing of the hybrid system components for a specied number of simulations ( ) over a scheduling period ( ). The optimal conguration of the hybrid power system ensures that the system reliability require- ments are met with minimum investment cost. F. Sensitivity Analysis The proposed methodology minimizes the investment cost of the hybrid power system for a specied level of reliability. Sensitivity analysis of the cost with respect to the reliability level helps the investors make a rational decision which is a compromise between the cost and reliability. An epsilon- constraint approach is used to determine the tradeoff region for the hybrid power system, where the cost minimization is the objective function and the reliability criteria are the constraints. Toward this end, is changed by values and the objective function is calculated for the desired reliability level. An efcient decision-making method is then required to realize the best compromise between the reliability and cost for the hybrid power system. We use a compromise solution that minimizes the normalized Euclidian distance between the po- tential optimal and utopia points. The utopia point is the point that is close enough to the minimum point. We solve the minimization [31] where is the normalized function for each decision criterion and is given by where and are the maximum and minimum values of and is the normalized utopia point, respectively. By solving (35) with , each criterion would be as close as possible to its minimum point. Normalization is necessary because different criteria have different units. III. CASE STUDIES Two cases are evaluated in this section for an optimal design of the hybrid power system. The rst case considers no load interruption and provides the sizing requirements of the hybrid power system for worst-case scenarios. It also gives the proba- bility distributions of the output random variables. Based on the design parameters, it then calculates the reliability index of the hybrid power system. The second case provides a stochastic framework to optimally satisfy the system reliability require- ments with the minimum investment cost. This scenario exam- ines the load shifting strategy and its potential impacts on the hybrid system reliability/cost evaluation. The wind speed, clear- ness index, and load data are from the Bonneville Power Administration authority and Mesonet [32], [33]. Capital costs for energy and power of the storage system are 0.2 $/Wh and 0.25 $/W, respectively [24]. The capital costs for wind and PV generation and the values of the system parameters for the case studies are provided in Table I. The parameters for the wind, PV, and load ARMA models are provided in Table II. A. Case I: Hybrid Power System Design Without the Reliability Constraints Case I uses the PS optimization to size the hybrid power system components for each simulation with no load interrup- tion. Fig. 4 shows the PDFs for the installed storage, wind, and PV capacities as well as the storage power and cost for 30% load shifting. These capacities fully meet the hybrid system load without the need to interrupt the load even for the worst-case scenarios of insufcient wind and PVgenerations. However, this design may result in a high investment cost for the hybrid power system. Using the expected values of these capacities avoids cost overestimation. Fig. 5 shows the expected values of the storage capacity and EIR for 10 000 simulations. The convergence state is reached after about 6000 simulations. The EIR is calculated for the scheduling period based on the design parameters to evaluate the reliability of the hybrid system. This gure shows the extent Fig. 3. Flowchart for the PS-based SMCS optimization. 368 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 2, APRIL 2014 to which the reliability requirements are met using the expected values of the PDFs for the installed storage, wind, and PV capacities. This design avoids the cost overestimation associated with designing the hybrid power system based on the maximum values of these PDFs. However, the reliability level may not seem appropriate for the hybrid power system. Designing the system based on the cost distribution is a postoptimization process which may not necessarily satisfy the desired reliability level. The cost distribution design is based on trial and error with no attempt at optimizing the parameter values. This demonstrates the necessity of hybrid system design with the reliability con- straints, which is the subject of investigation for Case II. B. Case II: Hybrid Power System Design With the Reliability Constraints Case II uses the proposed method to optimally size the hybrid system components while satisfying the desired reliability level for the hybrid power system. We evaluate the potential impacts of reliability level on the component sizing. The reliability constraint for EENS is set at a desired level and the hybrid system component-sizing problem is solved to minimize the investment cost and meet the desired level. Table III shows the wind, PV, and storage capacity installations and power for different reliability levels. The optimal costs for renewable energy generation and storage systemare also given in this table. As shown in Table III, increasing the desired reliability level ( ) from98.18%to 99.94%increases the cost of the hybrid power systemfor both the energy storage and renewable genera- tion. The optimum solutions show no PV installation for lower reliability levels while the wind installation is always contained within the optimal solutions. This is because wind energy, when available, can be generated for most periods of the day, whereas PV energy is generated during the daylight periods. Thus, more storage capacity is needed to provide energy during the hours when no solar energy is available. In addition, the installation cost of wind energy is less than the installation cost for PV generation. The increased variability of higher wind generation requires more storage capacity to deal with the uncertainties and sustains the higher reliability levels. This increases the cost of the hybrid wind and storage system. Therefore, the optimal design of the hybrid power systemfor higher reliability levels requires both the wind and PV generation to be installed within the system. This is due to the less uctuations in PV generation which decreases the storage capacity and cost of the hybrid wind, PV, and storage system when compared to the hybrid wind and storage system. The higher investment costs for higher reliability levels can be offset by anappropriate load shifting strategy for the hybrid power system. This is even more pronounced for HVAC loads which can be deferred and adjusted according to the renewable energy variations. Table IV shows the installed wind, PV, and storage capacities for energy index reliability of 99.39% ( ) and different load shifting percentages. As shown in Table IV, increasing the load shifting from 0% to 50% decreases the renewable and storage capacity installations for the hybrid power system. TABLE I INSTALLATION COSTS OF WIND AND PV GENERATION AND VALUES OF THE SYSTEM PARAMETERS [3], [11], [22], [23], [34], [35] TABLE II WIND, PV, AND LOAD ARMA PARAMETERS Fig. 4. PDFs for the installed storage, wind andPVcapacities, storage power, and cost of the hybrid power system. Fig. 5. Expected values of the storage capacity and EIR. ARABALI et al.: STOCHASTIC PERFORMANCE ASSESSMENT AND SIZING FOR A HYBRID POWER SYSTEM OF SOLAR/WIND/ENERGY STORAGE 369 The epsilon-constraint approach is used to determine the compromise between the cost and reliability (EENS), as shown in Fig. 6. Using least-squares curve tting, a curve is tted on the tradeoff points. Then, the best compromise is adopted as the nal solution utilizing the compromise-solution method. Table V shows the nal optimal plan for this tradeoff for different values of the utopia points close to 1 ( for cost and for EENS). Two decision criteria are considered for our problem: the cost ( ) and reliability index ( ). The maximum and minimum values of each criterion ( and ) are obtained from the tradeoff points shown in Fig. 6 to calculate the normalized function of (36). For a certain value of , increasing increases the reliability level (decreases EENS) for the hybrid power system. This is evident for higher values of , where the EENS decreases from0.0915 MWh for and to 0.0869 MWh for and . The same trend is observed for and , where the EENS decreases from 0.0964 to 0.0919 MWh and 0.1011 to 0.0966 MWh as increases from 0.95 to 1. Increasing the utopia values for reliability ( ) from 0.95 to 1 requires an investment cost that increases from 8.1853, 8.0783, and 7.9803 M$ to 8.2953, 8.1773, and 8.0733 M$ for , , and , respectively. Another observation is that increasing the utopia values for cost from 0.9 to 1 with the same decreases the reliability and cost of the system. IV. CONCLUSION This paper has proposed a stochastic framework for perfor- mance assessment of a hybrid power system to supply a deferra- ble HVAC load. The ARMA models are used to characterize the uncertainties of PV and wind power as well as HVAC loads. The PS-based SMCS optimization is proposed to optimally size the hybrid system components and satisfy the reliability require- ments. Optimal design of the hybrid power system is calculated for a residential feeder under different case studies. The simula- tion results for lower reliability levels showwind generation with no PVinstallation in the optimal solutions. The optimal design of the hybrid power systemfor higher reliability levels requires both wind and PVgeneration. Aload shifting strategy is developed to offset the higher investment costs for higher reliability levels. The simulation results showthe tradeoff between the systemcost and reliability for a hybrid power system. The best compromise is realized using a compromise-solution method. The decision- making process is performed for different values of the utopia points to determine the nal optimal plan based on the decisions maker preferences. The proposed methodology enables electric utilities to efciently utilize the installed renewable resources and storage capacities for supplying the controllable exible load, such as the HVAC load. APPENDIX REFERENCES [1] REN21 Steering Committee. (2012). Renewables 2012 Global Status Report. Paris, France REN21,[Online]. Available: http://www.map.ren21.net/GSR/ GSR2012_low.pdf. [2] J. P. Barton and D. G. Ineld, Energy storage and its use with intermittent renewable energy, IEEE Trans. Energy Convers., vol. 19, no. 2, pp. 441448, Jun. 2004. [3] G. Tina, S. Gagliano, and S. Raiti, Hybrid solar/wind power system probabilistic modeling for long-term performance assessment, Sol. Energy, vol. 80, no. 5, pp. 578588, May 2006. [4] L. Wang and C. Singh, Multicriteria design of hybrid power generation systems based on a modied particle swarm optimization TABLE III WIND, PV, AND STORAGE CAPACITY INSTALLATIONS FOR 30% LOAD SHIFTING AND DIFFERENT RELIABILITY LEVELS TABLE IV WIND, PV, AND STORAGE CAPACITY INSTALLATIONS FOR EIR OF 99.39% AND DIFFERENT LOAD SHIFTING PERCENTAGES Fig. 6. Tradeoff points for the compromise solution of the hybrid system. TABLE V FINAL DECISION MAKING FOR THE HYBRID POWER SYSTEM TABLE VI SIMULATION RESULTS FOR PS- AND GA-BASED SMCS 370 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 2, APRIL 2014 algorithm, IEEE Trans. Energy Convers., vol. 24, no. 1, pp. 163172, Mar. 2009. [5] C. Wang and M. H. Nehrir, Power management of a stand-alone wind/ photovoltaic/fuel cell energy system, IEEE Trans. Energy Convers., vol. 23, no. 3, pp. 957967, Sep. 2008. [6] O. C. Onar, M. Uzunoglu, and M. S. Alam, Modeling, control and simulation for an autonomous wind turbine/photovoltaic/fuel cell/ultra- capacitor hybrid power system, J. Power Sources, vol. 185, no. 2, pp. 12731283, Dec. 2008. [7] E. I. Vrettos and S. A. Papathanassiou, Operating policy and optimal sizing of a high penetration RES-BESS system for small isolated grids, IEEE Trans. Energy Convers., vol. 26, no. 3, pp. 744756, Sep. 2011. [8] A. Arabali, M. Ghofrani, M. Etezadi-Amoli, M. S. Fadali, and Y. Baghzouz, Genetic-algorithm-based optimization approach for energy management, IEEE Trans. Power Del., vol. 28, no. 1, pp. 162170, Jan. 2013. [9] H. Yang, W. Zhou, L. Lu, and Z. Fang, Optimal sizing method for stand- alone hybrid solar-wind system with LPSL technology by using genetic algorithm, Sol. Energy, vol. 82, no. 4, pp. 354367, Apr. 2008. [10] S. H. Karaki, R. B. Chedid, and R. Ramadan, Probabilistic performance assessments of autonomous solar-wind energy conversion systems, IEEE Trans. Energy Convers., vol. 14, no. 3, pp. 766772, Sep. 1999. [11] L. N. Kishore and E. Fernandez, Reliability well-being assessment of PV- wind hybrid system using Monte Carlo simulation, in Proc. Int. Conf. Emerging Trends in Electrical and Computer Technology (ICETECT), Roorkee, India, Mar. 2011. [12] R. Karki and R. Billinton, Reliability/cost implications of PV and wind energy utilization in small isolated power systems, IEEE Trans. Energy Convers., vol. 16, no. 4, pp. 368373, Dec. 2001. [13] A. Burgio, D. Menniti, A. Pinnarelli, and N. Sorrentino, Reliability studies of a PV-WG hybrid system in presence of multi-micro storage systems, in Proc. IEEE PowerTech Conf., Bucharest, Romania, Jun. 2009. [14] D. K. Khatod, V. Pant, and J. Sharma, Analytical approach for well-being assessment of small autonomous power systems with solar and wind energy sources, IEEE Trans. Energy Convers., vol. 25, no. 2, pp. 535545, Jun. 2010. [15] Cool Share for Southern Nevada [Online]. Available: http://www. nvenergy.com/home/saveenergy/rebates/coolshare.cfm. [16] Sequential Monte Carlo Methods [Online]. Available: http://www.stats.ox. ac.uk/~doucet/samsi_course.html. [17] R. Karki, P. Hu, and R. Billinton, A simplied wind power generation model for reliability evaluation, IEEE Trans. Energy Convers., vol. 21, no. 2, pp. 533540, Jun. 2006. [18] L. Martin, L. F. Zarzalejo, J. Polo, A. Navarro, R. Marchante, and M. Cony, Prediction of global solar irradiance based on time series analysis: Appli- cation to solar thermal power plants energy production planning, Sol. Energy, vol. 84, no. 10, pp. 17721781, Oct. 2010. [19] R. Karki, D. Dhungana, and R. Billinton, An appropriate wind model for wind integrated power systems reliability evaluation considering wind speed correlations, Appl. Sci., vol. 3, no. 1, pp. 107121, Feb. 2013. [20] S. Roy, Market constrained optimal planning for wind energy conversion systems over multiple installation sites, IEEE Trans. Energy Convers., vol. 17, no. 1, pp. 124129, Mar. 2002. [21] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, R. Seethapathy, M. Assam, and S. Conti, Adequacy evaluation of distribution system including wind/ solar DG during different modes of operation, IEEE Trans. Power Syst., vol. 26, no. 4, pp. 19451952, Nov. 2011. [22] A Comparison of Power Inverters by OUTBACK, SMA, FRONIUS and XANTREX [Online]. Available: http://www.ecodirect.com/Solar-Power- Inverter-Comparison-s/247.htm. [23] Green World Investor [Online]. Available: http://www.greenworldinvestor. com/topics/solar-renewable-energy-greeninvest/solar-inverters/. [24] T. Mosher, Economic valuation of energy storage coupled with photovol- taic: Current technologies and future projections, M.Sc. dissertation, Dept. Aeronautics and Astronautics Massachusetts Inst. Technology, Cambridge, MA, USA, Jun. 2010. [25] A. B. Rodrigues and M. G. Da Silva, Probabilistic assessment of available transfer capability based on Monte Carlo method with sequential simula- tion, IEEE Trans. Power Syst., vol. 22, no. 1, pp. 484492, Feb. 2007. [26] R. A. Gonzalez-Fernandez andA. M. Leite da Silva, Reliability assessment of time-dependent systems via sequential cross-entropy Monte Carlo simulation, IEEE Trans. Power Syst., vol. 26, no. 4, pp. 23812389, Nov. 2011. [27] N. R. Godha, S. R. Deshmukh, and R. V. Dagade, Time sequential Monte Carlo simulation for evaluation of reliability indices of power distribution system, in Proc. 2012 IEEE Symp. Computers and Informatics (ISCI 2012), Penang, Malaysia, Feb. 2012. [28] J. S. Al-Sumait, A. K. Al-Othman, and J. K. Sykulski, Application of pattern search method to power system valve-point economic load dispatch, Electr. Power Energy Syst., vol. 29, no. 10, pp. 720730, Dec. 2007. [29] R. M. Lewis and V. Torczon, A globally convergent augmented Lagrang- ian pattern search algorithm for optimization with general constraints and simple bounds, SIAM J. Optimiz., vol. 12, no. 4, pp. 10751089, 2002. [30] T. Weise, Global Optimization AlgorithmsTheory and Application, 2nd ed., Free Software Foundation, Jun. 2009 [Online]. Available: http:// www.it-weise.de/projects/book.pdf. [31] R. T. Marler and J. S. Arora, Survey of multi-objective optimization methods for engineering, Struct. Multidiscip. Optimiz., vol. 26, no. 6, pp. 369395, Apr. 2004. [32] Bonneville Power Administration (BPA) [Online]. Available: http:// transmission.bpa.gov/business/operations/wind/. [33] ISU Agclimate Data information, [Online]. Available: http://mesonet. agron.iastate.edu/agclimate/info.phtml. [34] G. Barbose, N. Darghouth, R. Wiser, and J. Seel. (2011, Sep.). Tracking the Sun IV: An Historical Summary of the Installed Cost of Photovoltaics in the United States. Lawrence Berkeley National Laboratory, Berkley, CA, USA, [Online]. Available: http://www.gsreia.org/site/wp-content/uploads/ 2012/10/ Historical-Summary-of-PV-Instillation-2011.pdf. [35] R. Wiser and M. Bolinger, 2010 Wind technologies market report, U.S. Department of Energy, Jun. 2011. Amirsaman Arabali (S11) received the B.Sc. degree in electrical engineering from Semnan University, Semnan, Iran, and the M.Sc. degree from Sharif University of Technology, Tehran, Iran, in 2005 and 2010, respectively. He is currently pursuing the Ph.D. degree at the University of Nevada, Reno, NV, USA. His research interests include power systems operation and planning, power market, energy management, and renewable energy. Mahmoud Ghofrani (M11) received the B.Sc. degree in electrical engineering from Amir-Kabir University of Technology, Tehran, Iran, and the M.Sc. degree from University of Tehran, Tehran, Iran, in 2005 and 2008, respectively. He is currently an Assistant Professor at the University of Washington, Bothell, WA, USA. His research interests include power systems operation and planning, renewable energy, and large scale integration of wind power generation. Mehdi Etezadi-Amoli (LSM11) received the Ph.D. degree from New Mexico State University, Las Cruces, NM, USA, in 1974. From 1975 to 1979, he worked as an Assistant Professor of electrical engineering at NewMexico State and the University of NewMexico, Las Cruces, NM, USA. From 1979 to 1983, he worked as a Senior Protection Engineer at Arizona Public Service Company, Phoenix, AZ, USA. In 1983, he joined the faculty of the Electrical Engineering Department at the University of Nevada, Reno, NV, USA. His current interest is in power system protection, large-scale systems, fuzzy control, neural network applications, and renewable energy. Mohammed Sami Fadali (SM91) received the Ph.D. degree fromUniversity of Wyoming, Laramie, WY, USA, in 1980. From1983 to 1985, he was a Postdoctoral Fellowat Colorado State University, Fort Collins, CO, USA. In 1994, he was a Visiting Professor at Oakland University, Rochester, MI, USA. He is currently a Professor of electrical engineering at the University of Nevada, Reno, NV, USA. His research interests include robust control, fuzzy-logic control, renewable energy systems, random signals, and fault detection. ARABALI et al.: STOCHASTIC PERFORMANCE ASSESSMENT AND SIZING FOR A HYBRID POWER SYSTEM OF SOLAR/WIND/ENERGY STORAGE 371