DG Allocation and Sizing Based on Reliability Improvement by Means of Monte Carlo Simulation

Transcription

1 DG Allocation and Sizing Based on Reliability Improvement by Means of Monte Carlo Simulation R. Yousefian, H. Monsef School of ECE, University of Tehran Technical Report {r.yousefian, Abstract The amount of distributed generation (DG) is increasing in the last few years, and it is predicted that in the future, it will play significant role in electrical energy systems. Based on DGs size and location, they may have positive and negative effects on system s behaviors, especially on reliability. This paper proposes sequential Monte Carlo simulation to evaluate the reliability indices, and Genetic Algorithm to calculate optimal location and size of DG based on reliability indices, with respect to loss and voltage constraints. One of the main advantages of utilizing sequential Monte Carlo simulation in bulk electric system reliability analysis is the ability to provide reliability index probability distributions in addition to the expected values of these indices. Keywords- Distributed Generation; Reliability Improvement; Monte Carlo Simulation; Genetic Algorithm. I. INTRODUCTION The share of distributed generators (DGs) in power systems has been slowly increasing worldwide due to some reasons like: growing rate of the population and load demands, problems of new power-plant and transmission lines investments, government policy changes, and increased availability of small capacity generation technologies [1]. DG can be defined as a small electric power source usually connected to distribution networks or even in the consumers side of the meter. However, there is not consistent definition of DG. Distributed generations in regard to their characteristics, size, and location may have remarkable positive effects in distribution systems. Possibilities include reactive power compensation to achieve voltage control, transmission loss reduction, spinning reserve to support outages, and access to renewable energy and a positive environmental impact [2], [3]. One of the main impacts of distributed generations is improvement of reliability in distribution system. It is in the best interest of all players involved to allocate and calculate the size of DG in an optimal way, such that the system reliability would be improved and system losses and voltage constraints would be considered [4]. As a result, genetic algorithm is utilized for optimization problem. According to what mentioned, one of the main necessity in planning and scheduling for DG application in distribution systems is using best methods to study the impact of these units on system reliability. Generally, there are two main methods to calculate the system reliability: analytical method and simulation technique. One of them is chosen in regard to system requirements. In this paper, a sequential Monte Carlo simulation technique is utilized to evaluate the reliability worth of the distribution system including DG. Conventional reliability evaluation using analytical methods only evaluate expected or average value of reliability indices. The actual shapes of the distribution of these indices can be achieved by means of Monte Carlo simulation as stochastic technique [5]. II. IMPACT OF DISTRIBUTED GENERATIONS ON RELIABILITY Distributed generations have made significant changes to power system structure. The distribution is traditionally passive and designed to operate with unidirectional energy flow from the source (transmission system) to the loads. In presence of DGs, the distribution system becomes an active system with both energy generation and consumption [1], [6]. Therefore, installing DG in distribution system may have significant impact on transmission power, voltages and reliability indices, which could be positive or negative. It would be positive if they are correctly coordinated with the rest of the network. DG s two main applications are illustrated bellow: A. Peak Shaving One of the DG applications, that is gaining popularity is the injection of power into the network when the main generation capacity in some feeders are lower than feeder s demand in peak time. DGs can support part of the demand, so feeder would be still in its limitation. Distributed generations as main or emergency generations improve both types of reliability indices of duration and number of failures. Interruption occurs when both generations are failed. B. Generation back-up A common example of DG usage is known as generation back up, in which the unit operates after failure of main generation in order to serve critical load points. After occurring failure in one of the elements, some of load points would be disconnected from the main generation source. At these times, DG can serve these load points in isolated part. This application of DG, also called intentional islanding, is the ability of managing an un-faulted part of the network by using DG, and plays a key role in active networks [7]. In this function of DGs, they merely improve reliability indices of failure s duration. In addition, any fault in main generation

2 causes interruption. In this case, DG increases restoration capacity of system. III. MONTE CARLO SIMULATION AND ITS APPLICATION IN RELIABILITY ANALYSIS Due to the random behavior of each components of the system, final result of the system, which is effected by total components, will be random [8]-[10]. In a real experiment of system, the occurrence is influenced by intrinsic behaviors of factors and variables of the system. While, in simulation, events are affected by models and probability distribution types used to display variables and components. The technique used in this paper is sequential simulation that can comprehensively take into account the chronological behavior of the system. The method, therefore, treats the problem as a series of experiments. There are advantages and disadvantages in both methods. Generally, Monte Carlo simulation requires a large amount of computing time compared to analytical methods [2]. A significant advantage of utilizing sequential Monte Carlo simulation in bulk electric system reliability analysis is its ability to provide reliability index probability distributions in addition to the expected values of these indices. Reliability index probability distribution analysis and its utilization are relatively new concepts in composite power system reliability analysis and decision making. There is frequently a need to know the range of a predictive reliability index and the likelihood of a certain value being exceeded [11]. IV. STEPS OF MONTE CARLO SIMULATION The procedure of Monte Carlo simulation is explained in following. For simulation, the most difficult problem is to find the load points influenced by the failure of an element and to calculate the restoration time, which is dependent on the network configuration. Failures in elements of system may affect one or more load points. 1- The initial state of each element is specified. Generally, it is assumed that all elements are initially in operating state. 2- For each element a random number is produced. These numbers have distributed uniformly between 0 and The duration of each component remaining in its present state (T on) is sampled from its probability distribution. Generally, the probability distribution function of T on is considered exponentially. 4- Comparing between the total T on of components, a component that has the least amount of T on is found. This element is the first element that failed. 5- The closest protective elements to the failure point isolate that element from other parts of the network. The failure time to the end of switching time is considered T sw. A uniform random number is produced for protective component, and with using the appropriate distribution function T sw is produced. 6- A uniform random number is produced for failed component, and with using the appropriate distribution function T off is produced. This is a down time for elements that could not be restored by protective elements. T on = 1 λ.ln(u) (1) T off = 1 μ.ln(u) (2) Where U and Ú are uniform random numbers, λ and µ are failure and repair rates of the elements. In Fig.1 operating and repairing cycle calculated with equations 1 and 2 is illustrated. Fig. 1. Operating and repairing time 7- Up and down times and total number of failure points are calculated and then return to stage1. 8- After calculating up and down times of load points, parameters λ, r and U and also SAIFI, CAIDI,... could be calculated as follows: λ i = (3) T ONi r i = T OFFi (4) SAIFI= λ i (5) CAIDI= U i μ i (6) ENS= L a(i) U i Where λ i is the failure rate, is the number of customers at load point i, U i is the annual outage time, and L a(i) is the average load connected to load point i. 9- After each repetition, simulation stopping rules can be used in order to extract acceptable simulation results [11], [12]. (7)

3 V. APPLICATION OF DISTRIBUTED GENERATIONS BASED ON RELIABILITY IMPROVEMENT The operation often leads to optimization methods. These methods search those groups of plans that help planner to achieve the predetermined goals. One of these frequently used methods in power systems is Genetic Algorithm. As was mentioned earlier, if distributed generation units are used for improving reliability, planning and operating problems are increased. In feeders with presence of DG, the optimization is not as simple as in the case of a feeder without DG. Because of the presence of additional generators, some portions of the feeder load may be satisfied after fault has been isolated [13], [14]. Applying distributed generation units in distribution networks makes power injected into buses changes. Therefore, current flows will change and causes changes in loss rate in networks. Power flow calculations can provide one of the most basic tools needed to calculate the criteria mentioned above. In this paper all simulations have been performed using Matpower package and all necessary constraints on computing power flow has been considered. Great attention should be paid toward the problem of allocating and sizing of DG. The installation of DG units at non-optimal locations results in more system losses and costs. Consequently, the development of an optimization methodology that could calculate the DGs allocation and size, improves system operation characteristics. The optimal DG allocation and sizing methodology should be able to provide the optimal solution that maximizes the benefits and minimizes the costs of DG installation. The methodology presented in this paper aims to optimize the allocation and size of DG in order to improve reliability level and to guarantee acceptable level of distribution network losses and voltage profile. VI. CASE STUDY In order to apply the method, distribution reliability test system of DIgSILENT Software (RTS) has been considered [15]. Fig.5 demonstrates the single line diagram of this system. This test system consists of 85 buses and 78 distribution lines. Total demand is 52.8 MW. The system provides the possibility of load transfer in case of contingencies. DG installation is considered to be possible in 11kV buses. In order to consider the effect of DGs, primary DGs of RTS are not included in case study. Monte Carlo simulation is carried out in distribution system. Reliability indices (failure rate and duration) for all load points are calculated through Eqs By means of these indices and through Eqs. 5-7 other indices are achieved. The results are demonstrated in Table I. TABLE I. RELIABILITY INDICES OF TEST SYSTEM WITHOUT DGS Total Reliability indices of Test System ENS (MWhr) SAIFI (1/yr) CAIDI (hr/yr) System s power flow is also analyzed and primary voltage ranges and total loss of the network is calculated through MatPower. The results are as shown in Table II. TABLE II. POWER FLOW OF TEST SYSTEM WITHOUT DGS Power Flow Voltage Range(Pu) Loss(MW) With utilizing the Monte Carlo simulation, also distribution of reliability indices can be calculated. That is one of the main advantages of Monte Carlo simulation. In Fig.2 distribution of CAIDI, SAIDI, and ENS of test system without DG is illustrated: Fig. 2. Distribution of system reliability indices without DGs For speeding up the calculation some buses that has high reliability indices could be eliminated. So the number of buses is limited to 20. Candidate locations for installing DGs are achieved by comparing the reliability indices of all load points, in this case ENS. The number of DGs that could be installed in distribution system is considered to be limited to 10. In Table.III indices of 20 worst buses regarding their ENS indices are listed. The first column is the substation number and the second column is the number of load point in corresponding substation. In Fig.3, ENSs of all buses are illustrated. The buses over the crucial lines are considered as critical buses and candidates for locating DGs. Fig. 3. ENS of all buses

4 Substati on number TABLE III. RELIABILITY INDICES OF CRITICAL LOAD POINTS Bus number ENSi (MWhr) λ (1/yr) r (hr) DGs location DGs capacity(mw) As could be seen in Table V and Table VI reliability indices and also loss and voltage levels are improved. The amount of loss reduced to 61.8% of its initial amount. TABLE V RELIABILITY INDICESOF TEST SYSTEM WITH DGS Total Reliability indices of Test System ENS (MWhr) SAIFI (1/yr) CAIDI (hr/yr) TABLE VI POWER FLOW OF TEST SYSTEM WITH DGS Power Flow Voltage Range(Pu) Loss(MW) In conclusion, the probability distributions of the reliability indices shown in Figs. 2 and 4 provide a representation of the manner in which the parameters vary around their mean values. Comparing probability distribution of SAIDI of system with DGs and without DGs shows that both distributions have unique characteristics means that DGs have no influence on occurring the failure. In contrast, probability distribution of CAIDI and ENS of system with DGs and without DGs shows that distributions of system with DGs have mean values closer to the ordinate axis. In other words, these indices have been improved noticeably due to improvement of restoring time. This function as mentioned above is called back-up generation. In order to determine the locations and sizes of DGs that pass the loss and voltage boundaries GA is applied through Matpower [16]. Limitation of voltage is considered 0.95 to 1.05 p.u, and limitation of loss is considered 6% of total MW injected to the distribution system. The results are shown in Table IV. TABLE IV. LOCATION AND SIZE OF DGS CALCULATED WITH GA DGs location DGs capacity(mw) Substation number Bus number Fig. 4. Distribution of system reliability indices with DGs VII. CONCULSION Size and location of DG are fundamental parameters in the application of DGs for improvement the reliability of system. This paper presents a method to calculate the optimum location for placement of definite DG units and size of them at different

5 buses, especially critical load points that have high ENS i. In each calculation loss and voltage limitation are considered too. In addition, this paper proposes Monte Carlo simulation to evaluate reliability, and Genetic Algorithm to identify the best location corresponding to the optimum size for improving total reliability. The advantage of the proposed methodology is that due to Monte Carlo simulation, distribution of reliability indices could be also calculated in addition to expected values. REFERENCES [1] D. H. Popovic, J. A. Greatbanks, M. Begovic, and A. Pregelj, Placement of distributed generators and reclosers for distribution network security and reliability, International Journal of Electrical Power & Energy Systems, Vol. 27, pp , [2] C. L. T. Borges and D. M. Falcao, Optimal distributed generation allocation for reliability, losses, and voltage improvement, International Journal of Electrical Power & Energy Systems, Vol. 28,pp , [3] Davoudi, M.G.; Bashian, A.; Ebadi, J., "Effects of unsymmetrical power transmission system on the voltage balance and power flow capacity of the lines," Environment and Electrical Engineering (EEEIC), th International Conference on, vol., no., pp.860,863, May 2012 [4] Moghadasi, S.-M.; Kazemi, A.; Fotuhi-Firuzabad, M.; Edris, A.-A., "Composite System Reliability Assessment Incorporating an Interline Power-Flow Controller," Power Delivery, IEEE Transactions on, vol.23, no.2, pp.1191,1199, April 2008 [5] Mohseni, A.; Mohajer Yami, S.; Akmal, A.A.S., "Sensitivity analysis and stochastic approach in study of transient recovery voltage with presence of superconducting FCL," Electrical Power and Energy Conference (EPEC), 2011 IEEE, vol., no., pp.479,484, 3-5 Oct [6] V. Van Thong, J. Driesen, and R. Belmans, Power Quality And Voltage Stability Of Distribution System With Distributed Energy Resources, International Journal of Distributed Energy Resource, Vol. 1, No. 3, pp , [7] F. Pilo, G. Celli, and S. Mocci, Improvement of reliability in active networks with intentional islanding, in Proc. of IEEE International Conference, Hong Kong, Vol. 2, pp , [8] Moghaddam, I.N.; Salami, Z.; Mohajeryami, S., "Generator excitation systems sensitivity analysis and their model parameter's reduction," Power Systems Conference (PSC), 2014 Clemson University, vol., no., pp.1,6, March 2014 [9] Moghaddam, I.N.; Salami, Z.; Easter, L., "Sensitivity Analysis of an Excitation System in Order to Simplify and Validate Dynamic Model Utilizing Plant Test Data," Industry Applications, IEEE Transactions on, vol.51, no.4, pp.3435,3441, July-Aug [10] Khosravani, S.; Naziri, I.; Afshar, A.; Karrari, M., "Fault tolerant control of Large Power Systems subject to sensor failure," Power Engineering, Energy and Electrical Drives (POWERENG), 2011 International Conference on, vol., no., pp.1,7, May 2011 [11] W. Wangdee, Bulk Electric System Reliability Simulation and Application, PHD Dissertation Dept. Electrical Eng., Univ. Saskatchewan Saskatoon, Dec [12] R. Billinton, Peng Wang, Teaching Distribution System Reliability Evaluation Using Monte Carlo Simulation, IEEE Trans. on Power Systems, Vol. 14, No. 2, May 1999 [13] A. Pregelj, M. Begovic,and A. Rohatgi, On Optimization of Reliability of Distributed Generation-Enhanced Feeders, in Proc. of International Conference on System Sciences, Hawaii, 2002 [14] S. Ehsani, Optimal Operation of Distributed Generation Considering Reliability Indices, Master Thesis Dept. Electric Eng.,Univ. Tehran, Iran, 2010 [15] Digsilent(2007), Power System Calculation Package, Available: [16] Davoudi, M.G.; Sadeh, J.; Kamyab, E., "Time domain fault location on transmission lines using genetic algorithm," Environment and Electrical Engineering (EEEIC), th International Conference on, vol., no., pp.1087,1092, May 20 Fig. 5. Test System