Determining the Optimal Capacity of Renewable Distributed Generation Using Restoration Methods Sung-Yul Kim, Student Member, IEEE, Wook-Won Kim, and Jin-O. Kim, Senior Member, IEEE AbstractThis paper proposes a methodology to determine the optimal capacity of renewable distributed generation (RDG). The objective in this paper is to maximize cost-savings of energy not supplied (ENS) as well as cost-savings of energy loss according to RDG installation in a distribution network. Additionally, in order to achieve this multi-objective optimization problem, both a restoration matrix and restoration availability are newly proposed. In this case study, a practical Korea Electric Power Corporation (KEPCO) system is presented and discussed, demonstrating the performance and effectiveness of the proposed methods. Index TermsEnergy loss, energy not supplied, optimal installed capacity, optimal interconnection bus, renewable dis- tributed generation, restoration availability, restoration matrix. NOMENCLATURE A. Abbreviations CIC Customer interruption cost. DG Distributed generation. ENS Energy not supplied. ESS Energy storage systems. EUE Expected unserved energy. IPP Independent power plant. ISO Independent system operator. KEPCO Korea Electric Power Corporation. KMA Korea Meteorological Administration. NPV Net present value. O&M Operation and maintenance. PF Power factor. PV Photovoltaic. RDG Renewable distributed generation. RES Renewable energy sources. Manuscript received July 30, 2012; revised January 13, 2013, April 27, 2013, July 11, 2013, and November 17, 2013; accepted January 13, 2014. Date of publication March 11, 2014; date of current version August 15, 2014. Paper no. TPWRS-00889-2012. S.-Y. Kim is with the Department of Energy Engineering, Keimyung Univer- sity, Daegu, South Korea (e-mail: energy@kmu.ac.kr). W.-W. Kimand J.-O. Kimare with the Department of Electrical Engineering, Hanyang University, Seoul, South Korea (e-mail: neocruser@hanyang.ac.kr; jokim@hanyang.ac.kr). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPWRS.2014.2305757 RPS Renewable portfolio standard. WT Wind turbine. B. Variables Coefcients of performance of WT . Power factor of WT . ct Customer type. Fault number in a contingency list. Number of RDG. Number of bus of customers. Number of bus interconnected with RDG. Repair time of a fault . Resistance of th element of Zbus matrix. Location of RDG installation. Observation time. th hour in a day. Interruption cost of each customer type during repair time . Cost of ENS before RDG installation. Cost of ENS for customers at bus after RDG installation. Total cost of ENS after RDG installation. Cost of energy loss before RDG installation. Total cost of energy loss after RDG installation. Net price for electricity. Total cost of ENS and energy loss before RDG installation. Optimal cost-saving of ENS and energy loss after RDG installation. Total cost of ENS and energy loss after RDG installation. Irradiation on location at observation time . Total number of bus in a network. Observation period (days). Total number of , and periods make a day. 0885-8950 2014 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. 2002 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 Active power injection in bus at time . Total active load of customers in bus at time . Active load of customer type ct in bus at time . Peak load of customer type ct at bus . Customer load pattern of customer type ct at time . Active power of RDG at time , which is installed in location and interconnected with bus . Installed capacity of RDG , which is installed in location and interconnected with bus . Maximum capacity of RDG in a network. Maximum capacity of RDG in bus . Output probability of RDG at , which is installed in location and interconnected with bus . Observed active power of PV at observation time , which is installed in location and interconnected with bus . Observed active power of WT in observation time , which is installed in location and interconnected with bus . Reactive power injection at bus at time . Total reactive load of customers at bus at time . Reactive power of RDG in time , which is installed in location and interconnected with bus . Observed reactive power of WT at observation time , which is installed in location and interconnected with bus . Restoration availability for a single RDG. Restoration availability for multiple RDG. Restoration matrix. U Unit function. Complex voltage at bus at time . Cut-in speed of WT. Cut-out speed of WT. Rated wind speed of WT. Wind speed on location at observation time . Inection point for the constant output of PV . Amplitude of branch admittance between bus and . Corresponding efciency of PV . Failure rate of a fault . Failure rate of customers at bus after RDG installation. Angle of branch admittance between bus and . C. Sets Set of RDG . Set of buses of customers . Set of interconnection buses for RDG. Set of installed locations for RDG. I. INTRODUCTION T HE technical improvements and the desire of customers for reliable and eco-friendly electric power have led to increased interest in distributed generation (DG) including re- newable energy sources (RES). Current energy policies of gov- ernments also help to promote installation and operation of this new type of generation through a feed-in tariff, quota system, carbon tax and trading, and tax relief. As a result, clean en- ergy technologies have become cost-competitive with conven- tional power systems, and in the near future, the generation cost of RES is expected to approach grid parity, which is the point at which alternative means of generating electricity produces power at a cost that is equal to or less than the price of pur- chasing power from the grid [1][4]. Nevertheless, at present the generation cost of RES is still more expensive than the gen- eration cost of conventional plants. Therefore, renewable energy projects are still suffering fromnancial hardship due to renew- able portfolio standard (RPS), which mandates that utilities and other load serving entities must procure a signicant portion of their customers electricity needs from RES [5], [6]. In this sit- uation, the planning of renewable distributed generation (RDG) must be carefully considered towards achieving sustainable hy- brid electric power systems. Various methodologies to facilitate prospective investors to arrive at an optimal planning for investment in RDG have been proposed [7][16]. This research allows evaluation and com- parison of total costs necessary in implementing the planning of RDG, considering construction costs, interconnection costs, operation and maintenance (O&M) costs, and generation rev- enue. Recently, an increasing concern for climate change has focused the interest of the public towards environmental pro- tection, thus [10], [11], and [15] presented an planning for in- vestment in RDGconsidering carbon emission. In [12][16], the net present value (NPV) analysis has been used for investment analysis, and the planning with the highest NPV was selected. The purpose of this research is to nd the best choice of new resources to be planned to maximize expected prots, from the perspective of independent power plant (IPP). On the other hand, there are many approaches for the plan- ning of RDG to minimize power loss and energy loss [17][22]. KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2003 Hung et al. [17] presented analytical expressions for nding size and location of DG. In this work, the objective was power loss minimization. Ochoa et al. [18] proposed the technique for ac- commodating RDG to minimize not power losses, but energy losses. Atwa et al. [19] proposed a planning technique for allo- cating various types of DG to minimize annual energy loss. The positive effects of RDG in these approaches are evaluated under normal network conditions without any faults in a network. As we consider an islanding condition in which RDG can continue to power a location independently even though electrical grid power is no longer present, customers load (that would oth- erwise have been left disconnected until repair had been com- pleted) can be transferred onto another part of the system such as RDG [23]. Therefore, this operational condition can have a remarkable effect on the distribution reliability such as energy not supplied (ENS). Above all previous approaches, however, overlook this positive effect of RDG when a fault occurs in a network. Dugan et al. [24] introduced the planning process for DG considering the value of expected unserved energy (EUE) as well as capacity constraints. Furthermore, [25] and [26] pro- posed the multi-objective optimization method for the planning of DG considering both ENS and energy loss. These previous works reected the positive effect of DG on reliability improve- ment. In the event of a system failure, however, it is hard to evaluate availability of electrical energy supplied from RDG to interrupted customers. Because network conguration, priority in power supply from RDG to customers, and the balance be- tween load and generation from RDG should be considered in order to evaluate this availability of RDG. Therefore, in this paper, restoration methods (restoration matrix and restoration availability) are newly proposed in order to solve this complex issue. This paper tries to improve on earlier works by using ENS in combination with energy loss according to the planning of RDG using the proposed restoration methods from the perspective of independent system operator (ISO). Here, the optimal installed capacity and the optimal bus interconnected with RDG is deter- mined by this proposed method. The proposed method for the planning of RDG permits a better simulation. The remainder of this paper is organized as follows: Section II presents the output modeling of RDG. The proposed restoration matrix and restoration availability are introduced in Section III, and the purpose of these restoration methods is to evaluate the reliability of a distribution network with RDG. Section IV describes the methodology for determining the optimal installed capacity and the optimal bus interconnected with RDG. Section V presents numerical results of application of the proposed methods in a practical KEPCO system, and relevant observations are also addressed in this section. Finally, conclusions and contributions to this paper are summarized in Section VI. II. RENEWABLE DISTRIBUTED GENERATION In general, DG is not centrally planned by the utility and not dispatched. The installed capacity of DG is normally smaller than 20 MW, and it is usually connected to a distribution system. DG can be classied according to various criteria, which are the purpose and the operation strategy, terminal characteristics, and the output controllability of DG [17], [27], [28]. Based on the output controllability of DG, coal-based generations and energy storage systems (ESS) are classied into the controllable unit type. On the other hand, the output of photovoltaic generation (PV) and wind turbines (WT) is intermittent and uncontrollable because these sources are affected by external factors such as location and weather. In this paper, PV and WT are considered for the output model of RDG. A. Output of Renewable Distributed Generation Based on the terminal characteristics, PV is regarded as the unit supplying only active power to the grid, and WT is re- garded as the unit supplying active power and consuming re- active power. 1) Only Active Power Supply: PV: Active power of PV is usually affected by irradiation, . The active power of PV in observation time can be represented as follows [29]: (1) where PV is installed in location , and it is interconnected with bus . In the relationship between irradiation and the output of PV, after a certain irradiation point, further increases in irra- diation produce relatively small changes in efciency. This in- ection point is represented as . 2) Active Power Supply and Reactive Power Consumption: WT: The output extracted by the rotor blades of a WT is the dif- ference between the upstreamand the downstreamwind powers. It varies with the density of the air sweeping the blades and with the cube of the wind speed [30]. For practical purposes, how- ever, active power of WT is commonly described as shown in (2) at the bottom of the page [31]. WT cannot export active power when the wind speed is below the cut-in speed , and it is shut down for safety reasons if the wind speed is higher than the cut-out speed . The output of WT is proportional to the velocity between the (2) 2004 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 cut-in speed and the rated wind speed . After this rated wind speed, the output of WT remains constant. Reactive power consumption of WT is represented as follows: (3) where is the power factor (PF) of WT . B. Output Model of Renewable Distributed Generation The output data of RDG is analyzed by using the output pa- rameters of RDG and the weather conditions during observation period, . Based on these data, the output model of RDG in time can be determined by using a linear rounding method, and the output probability of RDG is dened by the per-unit value as follows [32]: (4) (5) where is one among 24 hours in a day. Finally, the output of RDG in time can be represented in the following equations, where RDG is installed in location and interconnected with bus : (6) (7) In here, is zero while the type of RDG is PV. III. RESTORATION VIA RENEWABLE DISTRIBUTED GENERATION An islanding condition can have a remarkable effect on the reliability in a network with RDG. To evaluate this effect, there are a few issues that need to be considered including: 1) net- work conguration; 2) priority in power supply from RDG to customers; 3) the balance between load and generation from RDG. However, when a fault occurs in a network, it is hard to evaluate availability of electrical energy supplied from RDG to interrupted customers while simultaneously considering these issues. Furthermore, this priority in power supply differs from the conguration and the fault in a network, and also this bal- ance between load and generation depends on the network con- guration, the fault and time. In order to solve these compli- cated problems, restoration matrix and restoration availability are newly proposed in this paper. These restoration methods are innovative techniques for evaluating the distribution reliability compared to the method introduced in [5]. In this paper, these restoration methods are used to evaluate ENS for the optimal planning of RDG. Fig. 1. Simple distribution network with RDG. A. Restoration Matrix Let RDGbe interconnected with bus and a fault occur in a network. Then, , called the restoration matrix, offers valuable structural information about priority in power supply from RDG to customers at bus , and it is formulated as follows: if RDG cannot supply power to customers if customers are not affected by the fault priority. (8) Regardless of a fault in a network, if customers at bus can be supplied with power from the main grid, is zero. Additionally, zero also has the meaning of no cus- tomers at bus . On the other hand, if customers at bus cannot be supplied with power from the main grid or RDG due to net- work conguration during a fault , then is . For the rest, the value of the restoration matrix indicates a nat- ural number as priority in power supply from RDG to customers at bus . Fig. 1 depicts a simple case in order to explain restoration matrix. In here, RDG is interconnected with bus 6 , and customer and are located in bus 3, 5 and 7, re- spectively. When all 6 line faults are postulated in this network, restoration matrix can be determined in (9) by using denition (8): where row and column in restoration matrix are indicated by and , respectively. When a fault occurs on line 1 , RDG interconnected with bus 6 should consider to be a top priority (denoted by 1) in order to supply electricity. is second on the priority list (denoted by 2), followed by (denoted by 3) in this case KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2005 . Additionally, when a fault occurs on line 3 can be supplied with power from the main grid (denoted by 0), regardless of the output of RDG as well as the fault on line 3. , on the other hand, cannot be supplied with power from the main grid or RDG (denoted by ) due to this network conguration in Fig. 1 . B. Restoration Availability Using the restoration matrix, priority in the power supply is determined according to the faults in a network and buses in- terconnected with RDG. However, when a customer has the highest priority from power supply, it does not automatically mean the customer will be supplied with power during a fault in a network. This is because the balance between load and gener- ation is the basic rule of all power systems for a stable energy supply. Therefore, in order to evaluate availability of electrical energy supplied from RDG to interrupted customers, not only electricity supply priorities by restoration matrix, but the rela- tionship between load and generations in time is required. In this respect, restoration availability is proposed in addition to restoration matrix. , called restoration availability, offers valuable information about whether this power system maintains the bal- ance between customers load at bus and the output of RDG in time when a fault occurs in a network. Here, the output of RDG can be affected by the location for RDG installation, because weather conditions differ from location . In a network with a single RDG, restoration availability is determined as follows: (10) (11) (12) (13) (14) 0 means restoration via RDG is not available. 1, on the contrary, means restoration via RDG is available. Here, ct is a customer type such as public, commercial, industrial and resi- dential. is the set of bus with customers whose electrical load should be supplied from RDG during a fault . Active load at bus in time is , which is evaluated by using the load pattern of each customer type and the peak load of each customer type at bus . Here, is expressed as the per-unit value based on peak load of each customer type. As shown in (12), basically, restora- tion availability of RDG is determined by the balance between customers load and the output of RDG in time . Here, U is the unit function. Based on (10)(14), when various RDG are installed in a net- work, restoration availability is determined as follows: (15) where is the set of RDG . and are the set of installed locations for RDG and the set of interconnection bus for RDG, respectively. Restoration methods proposed in this paper are applied to evaluate the sum of cost-saving of ENS according to installed capacity and the bus interconnected with RDG in Section IV. IV. PROBLEM FORMULATION In this section, the method for determining the optimal in- stalled capacity and interconnected bus with RDG is proposed. In here, there are two different objectives. The rst objective is to minimize ENS when a fault occurs in a network (under abnormal network conditions), and the second one is to mini- mize energy loss without any fault (under normal network con- ditions). In this paper, these two objectives combined into a multi-objective optimization problem in order to evaluate all ef- fects of RDG installation. The purpose of this multi-objective optimization problem is to maximize the sum of cost-saving of ENS and cost-saving of energy loss, according to installed ca- pacity and the bus interconnected with RDG. The multi-objective function in this paper can be formulated as follows: (16) (17) (18) where is the sumof cost of ENS and cost of energy loss , before new RDG is installed in a network (denoted by Base Case). and are ENS cost and energy loss cost after RDG is installed in a network, respectively. The values of and are evaluated by considering all possible factors, which are all feasible RDG (in the set ), permitted locations (in the set ) and interconnectable buses (in the set ). The remainder of this section is composed of four parts. The rst and second part represent the way to evaluate using the proposed restoration methods and ac- cording to the planning of RDG, respectively. The third part describes constraints for this multi-objective optimization, and the last part depicts the ow chart of the proposed plan- ning technique. A. Cost of Energy Not Supplied: Customers are composed of various types, and customer in- terruption costs (CIC [$/kWh]) are different in each customer type and repair time. In this paper, CIC of customers at bus is evaluated by using CIC of each customer type and 2006 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 active load of each customer type . After RDG in- stallation, cost of ENS for customers at bus is determined as follows: (19) (20) (21) where and are repair time of a failure and the failure rate of a fault itself, respectively. is total number of , and periods make a day, e.g., hours a day. is the failure rate of customers at bus due to a fault when RDG (in the set is installed in location (in the set ) and is interconnected with bus (in the set ). is restoration availability which is evaluated by restoration methods in Section III. As a result, costs of ENS in the multi-objective optimization problem are determined as follows: (22) This equation indicates the sum cost of ENS for all customers in a network according to the planning of RDG. B. Energy Loss Cost: Power loss is generally used to investigate the impact of con- trollable energy sources. This approach is suitable for a snapshot at xed peak load and xed maximum output of DG. In order to investigate the time-varying demand as well as the intermittent RDG, however, energy loss is more reasonable for the planning of RDG [18], [33]. In this paper, therefore, energy loss is calcu- lated using variable demand and intermittent power driven by RDG in time . The total energy loss is commonly formulated as shown in (23)(25) at the bottom of the page [17], [34], where and are the active power injection at bus and the reactive power injection at bus in time , respectively. is the complex voltage at bus , and is the resistance of th element of Zbus matrix. is the total number of buses in a network. Fig. 2. Flow chart for the optimal planning of RDG. When RDG is interconnected with bus , RDG supplies with power to this bus . Customers at bus , on the contrary, use electricity. Therefore, the active/reactive power injection can be determined as follows: (26) (27) As mentioned in Section II, RDG may absorb or export reactive power depending on the terminal characteristics of RDG. If RDG is WT (the type of reactive power consumption), will be negative. (23) (24) (25) KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2007 Fig. 3. Cases system: Icheon and Jangheung in South Korea. As a result, energy loss costs in the multi-objective optimiza- tion problem are determined as follows: (28) where is the net price for electricity, and this equation indicates the sum of energy loss cost for one year because ENS is generally evaluated on a yearly basis. C. Constraints The multi-objective optimization problem should be mini- mized subject to certain equality and inequality constraints. Distribution power ow equations (29) (30) In these separating the active and reactive parts, and are the amplitude and angle of branch admittance between bus and bus [35]. Inequality constraints relate to keeping the magnitude of cur- rents at all lines and the magnitude of voltages at all buses within the permissible limits. Limits on line current magnitude (31) Limits on bus voltage magnitude (32) Limits on RDG capacity of interconnected bus (33) Limits on RDG capacity of a distribution network (34) where is maximum-allowable capacity of RDG in bus . For example, according to the South Korean grid code, maximum-allowable capacity of RDG was restricted to 20 MW in a distribution network. is maximum-allowable RDG capacity in a distribution network. D. Flow Chart The process for determining the optimal capacity of RDG is described as shown in Fig. 2. V. CASE STUDY In order to demonstrate the validity of the solution sets obtained by the proposed methods, a 154/22.9-kV distribution system is adopted as a case system, and this case system is a practical KEPCO (Korea Electric Power Corporation) system in the area of Icheon and Jangheung in South Korea. It is assumed that PV and WT, shown in Fig. 3, are installed in the adjacent area of this case system. According to the South Korean grid code for interconnection of RDG in a distribution network, maximum-allowable capacity of RDG in each inter- connection bus is as shown in Table I. Base Case shows the original network prior to the installation of RDG such as PV and WT. In Case 1, one of PV and WT is installed in the case system, and in Case 2, both of them are installed simultaneously. All system data and grid constraints are applied with KEPCO. Weather is still very stochastic and prone to uncertainty in terms of forecasts, even if various research studies have been performed [33], [36]. In this paper, the output of RDG at time is evaluated by using historical measurement data. The time- 2008 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 TABLE I MAX-ALLOWABLE CAPACITY OF RDG IN EACH BUS TABLE II CUSTOMER INTERRUPTION COST OF EACH CUSTOMER TYPE TABLE III CUSTOMERS PEAK LOAD based output models for PV and WT are derived from irradia- tion data and wind velocity data, respectively. These weather data have been acquired by KMA (Korea Meteorological Ad- ministration) during 2010, and depicted in Fig. 9 (Appendix). In this case study, an islanding operation of RDG is allowed in order to maximize the benets from RDG, because this oper- ation mode can improve the distribution reliability. The failure of lines, buses and transformers can be included in a contingency list. In this paper, the failure of all lines is pos- tulated for contingency, and it is assumed that these lines have a failure rate of 0.1 f/kmyr. Repair time in the main feeder ( ) is 4 hours, and repair time in the lateral distributors ( ) is 1 hour. Customer type (ct) is divided into public, commercial, indus- trial and residential in this paper. CIC ( depends on customer type and interruption time. CICof each customer type, shown in Table II, is bifurcated into 1 hour and 4 hours as inter- ruption time [37][40]. All customer load patterns in this case study are based on the investigation from 2003 to 2007 in South Korea. Customers peak load and load patterns are shown in Table III and Fig. 10, respectively. Using these data, and (13) and (14), hourly power demand of each customer is postulated. Energy loss is evaluated by the relation between customers load and the output of RDG in time as well as the grid data in the case system. Unit installed capacity of RDG is 1 kW for implementation of iterative technique, and the results in this case study are sim- ulated and evaluated by Matlab 7.0.4. All computer programs were run on a Windows 7-based PC (Intel Core i7 Q740 pro- cessor and 12 GB RAM). A. Analysis on Cost of ENS In Base Case where there are no RDG in the case system, cost of ENS in each customer is evaluated as shown in Table IV. TABLE IV BASE CASE: COST OF ENS TABLE V CASE 1: THE PLANNING OF RDG TO REDUCE COST OF ENS TABLE VI CASE 2: THE PLANNING OF RDG TO REDUCE COST OF ENS In order to evaluate ENS when RDG is installed and inter- connected to the case system, restoration matrix according to interconnected bus with RDG should be determined rst. Fig. 8 shows restoration matrix in this case study, which is the part of Matlab source code. In Case 1 where one of PV and WT is installed, lower cost of ENS compared to the results in Base Case are analyzed as shown in Table V. These results in Table V show that RDG installation does not always improve ENS. All buses interconnected with RDG but bus 2, 4 and 6 appear to have no effect on ENS, because ENS is affected by structural characteristics of a network (de- ned as restoration matrix) and the balance between customers load and the output of RDG in time (dened as restoration availability). In case of only PV installation, the most effective way to im- prove ENS is to install PV with 2974 kW (interconnection bus . In case of only WT installation, the most effective way to improve ENS is to install WT with 2824 kW (interconnection bus ). RDG is generally interconnected to the nearest bus for gener- ation expansion planning. In this case system, bus 4 for PV and bus 3 for WT are the nearest buses. However, when PV is in- terconnected to bus 6, this planning can reduce the cost of ENS more than 1.8% compared to the conventional planning where PV is interconnected to bus 4. Furthermore, to interconnect WT with the nearest bus 3 has no effect on ENS. In Case 2 where both of PV and WT are installed, various plans with two RDG can reduce costs of ENS compared to the results in Base Case. Table VI shows the top 3 plans among all feasible plans in Case 2. Although both of PV and WT are installed simultaneously in Case 2, all plans are not always able to generate cost-savings for ENS. Therefore, it is important to determine which bus we choose to interconnect, as well as the installed RDG capacity. KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2009 TABLE VII TOP 8 PLANS OF RDG TO IMPROVE COST-SAVING OF ENS TABLE VIII CASE 1: THE PLANNING OF RDG TO REDUCE ENERGY LOSS Considering all cases, the top 8 plans of RDG are arranged in order of improved cost-saving of ENS as shown in Table VII. With respect to improved cost-savings of ENS due to the planning of RDG, the most effective way is to install PV with 2974 kW (interconnection bus ) and WT with 2824 kW (interconnection bus ) simultaneously. Here, costs of ENS can be reduced by 4.39% compared to the results in Base Case, where there is no RDG in the case system. Additionally, this proposed planning can reduce the cost of ENS more than 2.8% compared to the conventional planning where RDG is intercon- nected to the nearest bus in a network. B. Analysis on Energy Loss Cost In Base Case, energy loss and energy loss cost are evalu- ated as 7830.6889 MWh and $783 068.9, respectively, by power ow. In Case 1, Table VIII shows the optimal capacity of RDG according to each bus , from a viewpoint of energy loss. In Table VIII, bus 1 for RDG interconnection does not inu- ence energy loss at all, because bus 1 is the secondary part of the main grid. From the results in Base Case and Table VIII, energy loss and cost-saving of energy loss in Case 1, can be depicted in Figs. 4 and 5. With respect to improved cost-saving of energy loss due to the planning of RDG, the most effective way is to interconnect RDG with bus 5. Here, two factors lead to this result. At rst, the electrical distance between and the main grid is relatively far. Secondly, customers load in is the largest in this case system as shown in Table III. When PV is interconnected with Fig. 4. Case 1(PV): energy loss and cost-saving of energy loss. Fig. 5. Case 1(WT): energy loss and cost-saving of energy loss. bus 5, this planning can reduce energy loss more than 1.5%com- pared to the conventional planning where PV is interconnected with the nearest bus 4. Additionally, when WT is interconnected with bus 5, this planning also can reduce energy loss more than 8.8% compared to the conventional planning where WT is in- terconnected with the nearest bus 3. Therefore, the planning of RDG which is interconnected with bus 5, is most effective to improve energy loss. In Case 2, Table IX shows the optimal capacity of RDG and interconnection bus , from a viewpoint of energy loss. Considering all Cases, the top 10 plans of RDG are arranged in order of improved cost-saving of energy loss as shown in Table X. With respect to improved cost-saving of energy loss due to the planning of RDG, the most effective way is to install PV with 3000 kW (interconnection bus ) and WT with 3000 kW (interconnection bus ) simultaneously. Here, cost of energy loss can be reduced by 40.7% compared to the results in Base Case, where there is no RDG in the case system. C. Optimal Planning of RDG The results in Case 1 are depicted in Figs. 6 and 7. As consid- ering only cost-saving of energy loss, the most effective way is 2010 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 TABLE IX CASE 2: THE PLANNING OF RDG TO REDUCE ENERGY LOSS TABLE X TOP 10 PLANS OF RDG TO IMPROVE COST-SAVING OF ENERGY LOSS Fig. 6. Case 1(PV): total improved cost-savings. to interconnect RDG with bus 5. However, this approach over- looks the effects of RDG when a fault occurs in a network. Therefore, the optimal planning of RDG proposed in this paper is determined by considering both sides of ENS and energy loss. As a result, the optimal planning of PV in Case 1 is to install PV with 3000 kW, and to interconnect PV with bus 6. On the other hand, the optimal planning of WT in Case 1 is to install WT with 3000 kW, and to interconnect WT with bus 4. By the proposed Fig. 7. Case 1(WT): total improved cost-savings. TABLE XI PROPOSED TOP 10 PLANS OF RDG method in this paper, the planning of PV can reduce annual cost of ENS and energy loss, $487 074.9, and the planning of WT can reduce them, $338 675.6. Finally, the top 10 plans of RDG are arranged in order of total improved cost-savings as shown in Table XI. The optimal planning of RDG in this case system is to install PV with 3000 kW (interconnection bus ) and WT with 3000 kW (interconnection bus ) simultaneously. Further- more, from this optimal planning, costs of ENS can be reduced by 4.39% and energy loss costs can be reduced by 39.12% com- pared to without RDGin this case system. This optimal planning of RDGis expected to achieve signicant annual cost-savings of $694 602. From the ISOs viewpoint, it is the maximum benet via RDG, considering under normal (energy loss) and abnormal conditions (ENS) using the proposed restoration methods. VI. CONCLUSION In this paper, a methodology to determine the optimal bus interconnected with RDG, as well as the optimal installed ca- pacity of RDG is proposed. Both of the ENS and energy loss are considered to solve this multi-objective problem. In here, restoration methods are also proposed to evaluate the effect on ENS according to the planning of RDG. KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2011 Fig. 8. Restoration matrix. Fig. 9. Weather data. (a) Irradiation in . (b) Wind speed in . In order to verify the proposed methods, data sets from real power systems in Icheon and Jangheung, South Korea are ap- plied to the case study. The results in the case study conrm that the proposed methods are more suitable to determine the op- timal planning of RDG compared to the conventional methods, because cost-savings of ENS under abnormal conditions as well as cost-savings of energy loss under normal conditions due to RDG installation are evaluated simultaneously. The planning of RDG in this paper is approached from a viewpoint of ISO, and this paper is innovative and fundamental towards sustainability and increasing the economic attractive- ness of RDG. In this sense, it can provide valuable insights to stakeholders in governments and companies. In this paper, the proposed restoration methods are only due to vicinity to RDG. In our future work, as considering the other factors in addition to vicinity, further developed methods will be proposed to improve its economic impacts and effectiveness. APPENDIX Figs. 810 showthe restoration matrix, weather data, and load patterns by customer type, respectively. 2012 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 29, NO. 5, SEPTEMBER 2014 Fig. 10. Load patterns by customer type. (a) Public. (b) Commercial. (c) Industry. (d) Residential. REFERENCES [1] J. Song and D. Kim, Trends of research and dissemination for PV systems in Korea, in Conf. Record 2006 IEEE 4th World Conf. Pho- tovoltaic Energy Conversion, 2006, vol. 2, pp. 22682271. [2] P. D. Lund, Fast market penetration of energy technologies in retro- spect with application to clean energy futures, Appl. Energy, vol. 87, no. 11, pp. 35753583, Nov. 2010. [3] M. Esteban and D. Leary, Current developments and future prospects of offshore wind and ocean energy, Appl. Energy, vol. 90, no. 1, pp. 128136, Feb. 2012. [4] D. Lewis, Solar grid parity[Power solar], Eng. Technol., vol. 4, no. 9, pp. 5053, 2009. [5] S.-Y. Kim and J.-O. Kim, Reliability evaluation of distribution network with DG considering the reliability of protective devices affected by SFCL, IEEE Trans. Appl. Supercond., vol. 21, no. 5, pp. 35613569, Oct. 2011. [6] S. Conti, Analysis of distribution network protection issues in pres- ence of dispersed generation, Elect. Power Syst. Res., vol. 79, no. 1, pp. 4956, Jan. 2009. [7] A. Mazer, Electric Power Planning for Regulated and Deregulated Markets. New York, NY, USA: Wiley-IEEE Press, May 2007. [8] C. De Jonghe, E. Delarue, R. Belmans, and W. Dhaeseleer, Deter- mining optimal electricity technology mix with high level of wind power penetration, Appl. Energy, vol. 88, no. 6, pp. 22312238, 2011. [9] A. I. Chatzimouratidis and P. A. Pilavachi, Technological, economic and sustainability evaluation of power plants using the analytic hier- archy process, Energy Policy, vol. 37, no. 3, pp. 778787, 2009. [10] Q. Chen, C. Kang, Q. Xia, and J. Zhong, Power generation expansion planning model towards low-carbon economy and its application in China, IEEE Trans. Power Syst., vol. 25, no. 2, pp. 11171125, May 2010. [11] J. Sirikum, A. Techanitisawad, and V. Kachitvichyanukul, A new ef- cient GA-benders decomposition method: For power generation ex- pansion planning with emission controls, IEEE Trans. Power Syst., vol. 22, no. 3, pp. 10921100, Aug. 2007. [12] F. Careri, C. Genesi, P. Marannino, M. Montagna, S. Rossi, and I. Siviero, Generation expansion planning in the age of green economy, IEEE Trans. Power Syst., vol. 26, no. 4, pp. 22142223, Nov. 2011. [13] S. Wong, K. Bhattacharya, and J. D. Fuller, Coordination of in- vestor-owned DG capacity growth in distribution systems, IEEE Trans. Power Syst., vol. 25, no. 3, pp. 13751383, Aug. 2010. [14] A. Kornelakis and E. Koutroulis, Methodology for the design op- timization and the economic analysis of grid-connected photovoltaic systems, IET Renew. Power Gener., vol. 3, no. 4, pp. 476492, 2009. [15] S. Wong, K. Bhattacharya, and J. D. Fuller, Long-term effects of feed-in tariffs and carbon taxes on distribution systems, IEEE Trans. Power Syst., vol. 25, no. 3, pp. 12411253, Aug. 2010. [16] N. Vaitheeswaran and R. Balasubramanian, Stochastic model for op- timal selection of DDG by Monte Carlo simulation, in Proc. Int. Conf. Power Elect. Drives and Energy Systems, 2006, pp. 15. [17] D. Q. Hung, N. Mithulananthan, and R. C. Bansal, Analytical expres- sions for DG allocation in primary distribution networks, IEEE Trans. Energy Convers., vol. 25, no. 3, pp. 814820, Sep. 2010. [18] L. F. Ochoa and G. P. Harrison, Minimizing energy losses: Optimal accommodation and smart operation of renewable distributed genera- tion, IEEE Trans. Power Syst., vol. 26, no. 1, pp. 198205, Feb. 2011. [19] Y. M. Atwa, E. F. El-Saadany, M. M. A. Salama, and R. Seethapathy, Optimal renewable resources mix for distribution system energy loss minimization, IEEE Trans. Power Syst., vol. 25, no. 1, pp. 360370, Feb. 2010. [20] A. Alarcon-Rodriguez, E. Haesen, G. Ault, J. Driesen, and R. Belmans, Multi-objective planning framework for stochastic and controllable distributed energy resources, IET Renew. Power Gener., vol. 3, no. 2, pp. 227238, 2009. KIM et al.: DETERMINING THE OPTIMAL CAPACITY OF RENEWABLE DISTRIBUTED GENERATION USING RESTORATION METHODS 2013 [21] V. H. M. Quezada, J. R. Abbad, and T. G. S. Romn, Assessment of energy distribution losses for increasing penetration of distributed generation, IEEE Trans. Power Syst., vol. 21, no. 2, pp. 553540, May 2006. [22] C. Wang and M. H. Nehrir, Analytical approaches for optimal place- ment of distributed generation sources in power systems, IEEE Trans. Power Syst., vol. 19, no. 4, pp. 20682076, Nov. 2004. [23] R. Billinton and R. N. Allan, Reliability Evaluation of Power Systems, 2nd ed. New York, NY, USA: Plenum, 1996. [24] R. C. Dugan, T. E. McDermott, and G. J. Ball, Planning for distributed generation, IEEE Ind. Applicat. Mag., vol. 7, no. 2, pp. 8088, 2001. [25] G. Celli, E. Ghiani, S. Mocci, and F. Pilo, A multiobjective evolu- tionary algorithm for the sizing and siting of distributed generation, IEEE Trans. Power Syst., vol. 20, no. 2, pp. 750757, May 2005. [26] M. F. Shaaban, Y. M. Atwa, and E. F. El-Saadany, A multi-objective approach for optimal DGallocation, in Proc. 2011 2nd Int. Conf. Elec- tric Power and Energy Conversion Systems (EPECS), 2011, pp. 17. [27] I.-S. Bae and J.-O. Kim, Reliability evaluation of distributed genera- tion based on operation mode, IEEE Trans. Power Syst., vol. 22, no. 2, pp. 785790, May 2007. [28] N. Jenkins, R. Allan, P. Crossley, D. Kirschen, and G. Strbac, Em- bedded generation, Inst. Eng. Technol., pp. 21132, Jun. 2000. [29] M. K. C. Marwali, H. Ma, S. M. Shahidehpour, and K. H. Abdul- Rahman, Short termgeneration scheduling in photovoltaic-utility grid with battery storage, IEEE Trans. Power Syst., vol. 13, no. 3, pp. 10571062, Aug. 1998. [30] M. R. Patel, Wind and Solar Power Systems: Design, Analysis, and Operation, 2nd ed. Boca Raton, FL, USA: CRC, Jul. 2005. [31] R. Billinton and G. Bai, Generating capacity adequacy associated with wind energy, IEEE Trans. Energy Convers., vol. 19, no. 3, pp. 641646, Sep. 2004. [32] J. Park, W. Liang, J. Choi, and J. Cha, A study on probabilistic reli- ability evaluation of power system considering solar cell generators, KIEE, vol. 58, no. 3, pp. 486495, Mar. 2009. [33] A. G. Marinopoulos, M. C. Alexiadis, and P. S. Dokopoulos, En- ergy losses in a distribution line with distributed generation based on stochastic power ow, Elect. Power Syst. Res., vol. 81, no. 10, pp. 19861994, Oct. 2011. [34] D. P. Kothari and J. S. Dhillon, Power System Optimization, 2nd ed. New Delhi, India: PHI Learning, 2010. [35] H. Saadat, Power SystemAnalysis, 2nd ed. NewYork, NY, USA: Mc- Graw-Hill, 2004. [36] W. El-Khattam, Y. G. Hegazy, and M. M. A. Salama, Investigating distributed generation systems performance using Monte Carlo simula- tion, IEEE Trans. Power Syst., vol. 21, no. 2, pp. 524532, May 2006. [37] S. B. Choi, K. Y. Nam, D. K. Kim, S. H. Jeong, H. S. Rhoo, and J. D. Lee, Assessment of interruption costs for residential customers, in Proc. KIEE 03 Conf., Nov. 2003, pp. 196198. [38] S. B. Choi, K. Y. Nam, D. K. Kim, S. H. Jeong, H. S. Rhoo, and J. D. Lee, Investigation and assessment of interruption costs for public customers in Korea, in Proc. KIEE 36th Conf. Part A, Jul. 2005, pp. 350352. [39] W.-I. Hwang, K.-B. Song, and B.-W. Nam, Model establishment for customer interruption cost estimate, in Proc. KIEE 06 Conf., Nov. 2006, pp. 174176. [40] S. B. Choi, Investigation and assessment of interruption costs for in- dustrial customers in Korea, KIEE Mag., vol. 56, no. 12, pp. 6472, Dec. 2007. Sung-Yul Kim (S11) received the B.S. and Ph.D. degrees in electrical en- gineering from Hanyang University, Seoul, South Korea, in 2007 and 2012, respectively. From 2012 to 2013, he was a research assistant at Georgia Institute of Tech- nology, Atlanta, GA, USA. Since 2013, he has been with the Department of Energy Engineering, Keimyung University, Daegu, South Korea. His main re- search interests include computer aided optimization, renewable energy sources applied to smart grid, and power system reliability. Wook-Won Kim received the B.S. and M.S. degrees in electrical engineering from Hanyang University, Seoul, South Korea, in 2008 and 2011, respectively. He is pursuing the Ph.D. degree in electrical engineering of Hanyang University. His research interests include power system reliability, optimal scheduling of energy storage system with heuristic algorithm, and forecasting renewable energy sources. Jin-O. Kim (SM03) received the B.S. and M.S. degrees in electrical engi- neering from Seoul National University, Seoul, South Korea, and the Ph.D. de- gree from Texas A&M University, College Station, TX, USA. He is presently a Professor with the Department of Electrical Engineering, Hanyang University, Seoul, South Korea. His research interests include power system reliability, planning, and power economics applied to smart grid.