Reiabiity Improvement with Optima Pacement of Distributed Generation in Distribution System N. Rugthaicharoencheep, T. Langtharthong Abstract This paper presents the optima pacement and sizing of distributed generation (DG) in a distribution system. The probem is to reiabiity improvement of distribution system with distributed generations. The technique empoyed to sove the minimization probem is based on a deveoped Tabu search agorithm and reiabiity worth anaysis. The deveoped methodoogy is tested with a distribution system of Roy Biinton Test System (RBTS) bus 2. It can be seen from the case study that distributed generation can reduce the customer interruption cost and therefore improve the reiabiity of the system. It is expected that our proposed method wi be utiized effectivey for distribution system operator. Keywords Distributed generation Optimization technique Reiabiity improvement, Distribution system. I. INTRODUCTION HE distribution system is an important part that provides Tthe fina ink between the utiity and customers. In practice, most distribution systems have a singe-circuit main feeder and are configured radiay. The radia distribution system is widey used because of its simpe design, generay ow cost and supportive protection scheme. This configuration suggests that a components between a oad point and the suppy point shoud competey operated and therefore poor reiabiity can be expected as the faiure of any singe component causes the oad points to be disconnected. An idea aternative on eectric distributions to eectric users is the instaation of a sma sized generator or commony known as distributed generation (DG) []. DG is a sma sized generator connected in parae with the distribution system. DG is expected to pay an increasing roe in emerging eectric power systems. Studies have predicted that DG wi be a significant percentage of a new generation going on ine. There are severa different types of resources and technoogies that can be used for DG such as wind, soar, fue ces, hydrogen, and biomass. DG can resut in a network operation and panning practices with economic impications. The benefits of DG are cassified into two groups: technica and economics [2]. For exampe, oss, votage profie, reiabiity of suppy, maintenance costs, and network connection reinforcement costs can be affected by the connection of DG to the distribution system [3], [4]. Eectric utiities, therefore, can benefit from the instaation of DG. N.Rugthaicharoencheep is with Department of Eectrica Engineering, Facuty of Engineering, Rajamangaa University of Technoogy Phra Nakhon, Bangkok, Thaiand (e-mai: nattachote.r@rmutp.ac.th). T. Langtharthong is with Department of Eectrica Engineering, Facuty of Engineering, Rajamangaa University of Technoogy Phra Nakhon, Bangkok, Thaiand (e-mai: thong.@rmutp.ac.th). Distribution system reiabiity assessment can, in genera, be divided into the two basic tasks of assessing past performance and predicting future performance. Predicting reiabiity performance is usuay concerned with the suppy adequacy at the customer oad points [5]. The conventiona approach to teaching distribution system reiabiity evauation, in either a university or industry based setting, is to use the basic anaytica equations to cacuate oad point faiure rates, average outage durations and average annua outage times. The technique empoyed to sove the minimization probem is based on a deveoped Tabu search agorithm and reiabiity worth anaysis. The Tabu agorithm systematicay searches soutions expressed in forms of the ocation and size of DGs. The soution obtained wi then be passed to reiabiity worth anaysis to evauate the quaity of the soution. The process is repeated unti the best soution has been found. The deveoped methodoogy is tested with a distribution system of Roy Biinton Test System (RBTS) bus 2. II. TABU SEARCH Tabu search is a meta-heuristic that guides a oca heuristic search strategy to expore the soution space beyond oca optimaity [6]. The basic idea behind the search is a move from a current soution to its neighborhood by effectivey utiizing a memory to provide an efficient search for optimaity. The memory is caed Tabu ist, which stores attributes of soutions. In the search process, the soutions are in the Tabu ist cannot be a candidate of the next iteration. As a resut, it heps inhibit choosing the same soution many times and avoid being trapped into cycing of the soutions [7]. The quaity of a move in soution space is assessed by aspiration criteria that provide a mechanism for overriding the Tabu ist. Aspiration criteria are anaogous to a fitness function of the genetic agorithm and the Bozman function in the simuated anneaing as shown in Fig.. In the search process, a move to the best soution in the neighborhood, athough its quaity is worse than the current soution, is aowed. This strategy heps escape from oca optima and expore wider in the search space. A Tabu ist incudes recenty seected soutions that are forbidden to prevent cycing. If the move is present in the Tabu ist, it is accepted ony if it has a better aspiration eve than the minima eve so far [8]. Fig. 2 shows the main concept of a search direction in Tabu search. 308
Average service unavaiabiity index (ASUI) ASUI = ASAI = UN i i 8760 (5) Energy not suppied index (ENS) ENS = Lai () Ui (6) Fig. Search direction of Tabu search Fig. 2 Search direction of Tabu search III. DISTRIBUTION SYSTEM RELIABILITY The basic distribution system reiabiity indices at a oad point are average faiure rate λ, average outage duration r, and annua outage duration U. With these three basic oad point indices, the foowing system reiabiity indices can be cacuated [8]. Average interruption frequency index (SAIFI) λ = in SAIFI i () System average interruption duration index (SAIDI) UN SAIDI = i i (2) Customer average interruption duration index (CAIDI) UN CAIDI = i i (3) λin i Average service avaiabiity index (ASAI) ASAI = 8760 Ui 8760 (4) Average energy not suppied index (AENS) AENS = Lai () Ui λ i = faiure rate of contingency i N i = tota number of oad points i U = i annua outage duration i L = ai () faiure rate of contingency i A basic approach to quantifying the worth of eectric service reiabiity is to estimate customer interruption costs due to eectric power suppy interruptions. One convenient way is an interpretation of customer interruption costs in terms of customer damage functions. The customer outage cost ( ECOST ) is cacuated from reiabiity indices of the oad point and customer damage function [9]-[0]. IV. PROBLEM FORMULATION The objective is to minimize the customer outage cost that can be written as foows: nh ni Minimize ECOST = ( Lai () Chirh λh ) (8) h= i= L = average oad connected to oad point i ai () C = hi outage cost ($/kw) of customer due to contingencyh λ h = faiure rate of contingency h r h = average outage time of contingency h n h = number of contingencies h n i = tota number of oad points i Constraints: ) Power fow equations: (7) Pk = YikVV i k cos( θik + δk δi) (9) i= Qk = YikVV i k sin( θik + δk δi) (0) i= 309
2) Votage of each bus k must be within specified imits: V min k Vk Vk () 3) Current transfer capabiity of feeders : I I ; {,2,..., N } (2) 4) Maximum number of DGs ( n DG ) to be instaed: ejk ndg j {, 2,..., N C } (3) k = 5) Maximum instaed capacity of DGs: N C C j e jk G (4) k= j= 6) Decision variabes for the instaation of a DG: ejk 0 if the DG is not instaed at bus k = if the DG is instaed at bus k with the capacity at step j (5) 7) Ony one DG can be instaed at one position N C e jk k {, 2,..., } (6) j = P k = power active power at bus k Q k = power reactive power at bus k = number of buses Y ik = eement ( i,k ) in bus admittance matrix θ ik = ange of Y ik δ k = votage ange at bus k N = number of feeders min V = k minimum votage at bus k V = k imum votage at bus k I = current fow in feeder I = imum current capabiity of feeder = number of buses N C = number of capacity steps of a DG C = capacity at step j of DG j G = imum tota instaed capacity N C = number of capacity steps of a DG C = capacity at step j of DG j e = decision variabe for instaation of a DG at bus k jk with the capacity at step j iteration and iteration index m=. Step 2: Let the initia soution obtained in step be the current soution and the best soution: S best = S 0, and S current = S 0. Step 3: Perform a power fow anaysis to determine whether the current soution satisfies the constraints defined in (9) and (0). A penaty factor is appied for constraint vioation. Step 4: CacuateECOST using (8) with consideration of oad point restoration. Step 5: Cacuate the aspiration eve of S best : f best = f(s best ). The aspiration eve is the sum of ECOST and a penaty function. Step 6: Generate a set of soutions in the neighborhood of S current. This set of soutions is designated as S neighbor. Step7: Cacuate the aspiration eve for each member of S neighbor, and choose the one that has the highest aspiration eve, S neighbor_best. Step 8: Check whether the attribute of the soution obtained in step 7 is in the Tabu ist. If yes, go to step 9, or ese S current = S neighbor_best and go to step 0. Step 9: Accept S neighbor_best if has a better aspiration eve than f best and set S current = S neighbor_best, or ese seect a next-best soution that is not in the Tabu ist to become the current soution. Step 0: Update the Tabu ist and set m = m+. Step : Repeat steps 6 to 0 unti the specified imum iteration has been reached and report the best soution. Summing up, the performance of Tabu search depends on a proper choice of the neighbor of a soution, on the number of iterations for which a move is kept as Tabu, on the best combination of short- and ong-term memory and on the best baances of intensification and diversification mechanism [0]. The soution agorithm for the probem is described step by step as shown in Fig. 3. VI. CASE STUDY The deveoped Tabu search agorithm is tested with a distribution system of RBTS bus 2 [] to minimize the customer outage cost. There are 4 feeders and 22 oad points. The peak oading eve of bus 2 is 20 MW. The configuration of the system is shown in Fig. 4. The imum iteration for Tabu search is 00. The minimum and imum votages for each bus are 0.95 p.u. and.05 p.u., respectivey. The sizes of DGs are 00 kw-,500 kw. The faiure of a transformer is recovered by repair. A protective devices and DGs are assumed to be fuy reiabe. Three cases are investigated in Tabe I. The resuts from the case study are shown in Tabes II and III. V. SOLUTION ALGORITHM The soution agorithm for the probem is described step by step as foows: Step : Randomy seect a feasibe soution from the search space: S 0 Ω. Set the size of a Tabu ist, imum 30
Case TABLE I CASE STUDY FOR RELIABILITY ANALYSIS Maximum number of DGs, n DG (unit) Tota instaed capacity, G (kw) - - 2 000 3 3 2000 4 3 3000 5 4 4000 Fig. 3 Fowchart for soution agorithm Fig. 4 RBTS bus 2 Radia Distribution system TABLE II OPTIMAL PLACEMENT AND SIZING OF DGS Capacity of DG instaed Tota capacity of DG (kw) Case Location of DG (bus) (kw) - - - 2 5 500 0500 3 3, 5, 00, 500, 500 00 4 0,, 5 000, 600, 200 2800 5 5, 0, 2, 5 600, 200, 200, 200 3200 A the cases have the same SAIFI because this index depends ony on the reiabiity of components (e.g., ines, transformers) and is not affected distributed generations to be instaed. We can see that the overa reiabiity indices of cases 2 to 5 in Tabe III are improved compared with that of case (base case). In cases 2, 3, 4 and 5, the number of DGs is imited at, 2, 3 and 4 unit respectivey, see reductions in the system ECOST. TABLE III RESULT OF CASE STUDY FOR RELIABILITY INDICES Reiabiity indices Cases 2 3 4 5 SAIFI (interruptions/customer yr) 0.2482 0.2482 0.2482 0.2482 0.2482 SAIDI (hours/customer yr) 3.732 3.7290 3.726 3.7253 3.725 CAIDI (hours/customer 5.036 5.024 5.02 5.009 5.008 interruption) ASAI 0.9996 0.9996 0.9996 0.9996 0.9996 ASUI 0.0004 0.0004 0.0004 0.0004 0.0004 ENS (kwh/year) 40,775.30 40,509.70 40,256.00 39,455.40 39,443.60 AENS (kwh/customer/year) 2.37 2.23 2.0 20.68 20.67 ECOST ($/year) 49,922.30 47,552.80 45,288.70 43,262.90 43,57.60 ECOST reduction (%) - 4.75 9.28 3.34 3.55 It is very interesting to note that the constraint given in (4) is binding for these four cases. The reason is that to minimize the system ECOST, as many DGs as possibe shoud be instaed. Distribution system of RBTS bus 2 with distributed generation for case 5 shown in Fig. 5 It is observed that a DG, if its size is arge enough, tends to be instaed at the end of a feeder. 3
[7] F. Gover, Tabu search-part I. ORSA J. Computing, vo., no. 3. 989. [8] M. Hiroyuki, and O. Yoshihiro, Parae tabu search for capacitor pacement in radia distribution system, in Proc. Power Engineering Society Winter Meeting, vo. 4, pp. 2334-2339, 2000. [9] L. Goe, and R. Biinton, Procedure for evauating interrupted energy assessment rates in an overa eectric power system, IEEE Trans. Power Systems, vo. 6, no. 4, pp.398-403, 99. [0] L. Goe, and R. Biinton,. Basic data and evauation of distribution system reiabiity worth, in Proc. Computer, Power and Communication Systems in a Rura Environment, pp. 27-277, 99. [] Y. J. Jeon, and J. C. Kim, Appication of simuated anneaing and tabu search for oss minimization in distribution systems, Eectrica Power and Energy Systems, vo. 26, no., pp. 9-8, 2004. Fig. 5 Distribution system of RBTS bus 2 with distributed generation for case 5 VII. CONCLUSION The search for the best compromise among the objectives is achieved by Tabu search technique for optima pacement and sizing of distributed generation in distribution systems. Empoying DG in a distribution system resuts in severa benefits such as increased overa system efficiency. The effectiveness of the proposed method was demonstrated by a case study of a distribution network of RBTS bus 2. It can be seen from the case study that distributed generation can reduce the customer interruption cost and therefore improve the reiabiity of the system. It is expected that our proposed method wi be utiized effectivey for distribution system operator. ACKNOWLEDGMENT The authors woud ike to express his gratitude to Rajamangaa University of Technoogy Phra Nakhon, Thaiand for support. REFERENCES [] D. T. Wang, L. F. Ochoa, and G. P. Harrison, DG impact on investment deferra: network panning and security of suppy, IEEE Trans. Power System, vo. 25, no. 2, pp. 34-4, May 200. [2] J. Zhang, H. Cheng, and C. Wang, Technica and economic impacts of active management on distribution network, Eectrica Power and Energy Systems, vo. 3, pp. 30-38, 2009. [3] J. Mutae, Benefits of active management of distribution networks with distributed generation, in Proc. Power System Conf. and Exposition. pp. 60 606, 2006. [4] L. F. Ochoa, A. P. Fetrin, and G. P. Harrison, Evauating distributed timevarying generation through a mutiobjective index, IEEE Trans. Power Deivery, vo. 23, no. 2, pp. 32-38, Apri 2008. [5] R. Biinton, and R. N. Aan, Reiabiity evauation of power systems, pitman advanced pubishing program, 984. [6] D. Bernaand, and A. Cigdem, Simuation optimization using tabu search, in Proc. Winter Simuation Conference, pp. 805-80. 2000. Nattachote Rugthaicharoencheep (M 0) received his Ph.D. in Eectrica Engineering from King Mongkut s University of Technoogy North Bangkok (KMUTNB), Thaiand in 200. He is currenty a ecturer at the Department of Eectrica Engineering, Facuty of Engineering Rajamangaa University of Technoogy Phra Nakhon (RMUTP), Bangkok, Thaiand. His research interests incude power system operation, optimization technique, and distributed generation. Thong Lantharthong received his M.Eng in Eectrica Engineering from Rajamangaa University of Technoogy Thanyaburi, Pathumthani, Thaiand in 200. He is currenty a assistant professor at the Department of Eectrica Engineering, Facuty of Engineering Rajamangaa University of Technoogy Phra Nakhon (RMUTP), Bangkok, Thaiand. His research interests incude power system operation, optimization technique, and distributed generation. 32