Research Article An Improved Genetic Algorithm for Power Losses Minimization using Distribution Network Reconfiguration Based on Re-rank Approach

Transcription

1 Research Journal of Appled Scences, Engneerng and Technology 8(8): , 2014 DOI: /raset ISSN: ; e-issn: Maxwell Scentfc Publcaton Corp. Submtted: May 21, 2014 Accepted: July 13, 2014 Publshed: August 25, 2014 Research Artcle An Improved Genetc Algorthm for Power Losses Mnmzaton usng Dstrbuton Network Reconfguraton Based on Re-rank Approach N.H. Shamsudn, N.F. Omar, A.R. Abdullah, M.F. Sulama, N.A. Abdullah and H.I. Jaafar Faculty of Electrcal Engneerng, Unverst Teknkal Malaysa Melaka, Hang Tuah Jaya, Duran Tunggal, Melaka, Malaysa Abstract: Ths study presents the mplementaton of Improved Genetc Algorthm (IGA) to mnmze the power losses n the dstrbuton network by mprovng selecton operator pertanng to the least losses generated from the algorthm. The maor part of power losses n electrcal power network was hghly contrbuted from the dstrbuton system. Thus, the need of restructurng the topologcal of dstrbuton network confguraton from ts prmary feeders should be consdered. The swtches dentfcaton wthn dfferent probabltes cases for reconfguraton purposes are comprehensvely mplemented through the proposed algorthm. The nvestgaton was conducted to test the proposed algorthm on the 33 radal busses system and found to gve the better results n mnmzng power losses and voltage profle. Keywords: Dstrbuton Network Reconfguraton (DNR), dstrbuton systems, Genetc Algorthm (GA), Improved Genetc Algorthm (IGA), power losses INTRODUCTION In 1992, Malaysa suffers from a total blackout due to the lghtnng that strkes onto the transmsson faclty and leads falure for both transmsson and dstrbuton system. Another maor dsturbance occurs on the transmsson lne at Terengganu trpped. The power system network trpped and caused blackout once agan n 1996 and consequently affected to Kuala Lumpur, Selangor, Putraaya, Johor, Melaka and Neger Semblan for several hours. After that n 2003, a power falure caused 5 h of blackout that affected southern parts of Pennsular Malaysa ncludng Malacca, Johor, Selangor, Neger Semblan and Kuala Lumpur. Then followed by 2005 the faults occurred at the man cable transmsson lne grd that caused blackout n northern pennsular ncludng Perak, Penang, Kedah and Perls. Nevertheless, numerous reports were brought up regardng blackout around Pennsular Malaysa, Sabah and Sarawak n These knds of crcumstances happened because the dstrbuton system tself suffers from overloaded, load varatons, malfuncton of equpment and damaged by thrd party or work qualty (Nza Samsudn, 2009). Hence, many algorthms researches have been conducted manly on conventonal optmzaton approaches, heurstc technques and artfcal ntellgence technques (Cheragh and Ramezanpour, 2012) wth the am of reducng the power losses at dstrbuton system. A dscrete branch and bound method was ntroduced n Cheragh and Ramezanpour (2012) n order to reduce power losses n dstrbuton network. It nvolved meshed network by ntally closng swtches n the network. Then, a new radal confguraton s reached by openng swtches one at a tme. Whle accordng to Hu et al. (2010), a modfed heurstc algorthm s ntroduced to restore the servce and load balance. Ths method s done through a modfed fast decoupled Newton-Raphson Method to check the operaton constrants. A study n Zhao et al. (2012) offered a heurstc reconfguraton based on branch exchange to reduce power losses and to ensure balance of loads n feeders. However, ths type of reconfguraton was very tme consumng and t does not smply solve complexes constrant problems. Whle a study conducted n Zhao et al. (2009) that s applcable n radal-type dstrbuton system presents optmzaton of radal dstrbuton systems usng an effcent algorthm by network reconfguraton and capactor allocaton. Through the usage of both capactor allocaton and dedcated genetc algorthm wth specal crossover and mutaton operators, real power losses at 69-bus test system and 135-bus test system are reduced. Hence, by acknowledgng the lmtaton of GA n solvng the losses mnmzaton of dstrbuton network, an mproved GA needs to be ntroduced. Ths study proposes the use of IGA to reduce the power losses n the dstrbuton network wth an Correspondng Author: N.H. Shamsudn, Faculty of Electrcal Engneerng, Unverst Teknkal Malaysa Melaka, Hang Tuah Jaya, Duran Tunggal, Melaka, Malaysa Ths work s lcensed under a Creatve Commons Attrbuton 4.0 Internatonal Lcense (URL:

2 Fg. 1: Radal 33-bus network after reconfguraton mprovement n selecton operator and the adustment of crossover and mutaton probabltes. The dfferent probabltes cases for reconfguraton purposes are conducted through the proposed algorthm for swtches dentfcaton. The IGA used re-rank approach to reduce the power losses and mprove voltage profle of a 33- bus radal dstrbuton network. Model descrpton: To demonstrate the effectveness of the proposed algorthm, the radal 33-busses system s used whch conssts of 33 nodes, 38 swtches where 5 of them are te swtches and the remanng 33 are sectonalzng swtches (Shamsudn et al., 2014; Sulama et al., 2013, 2014a, b). Fgure 1 llustrates the dstrbuton network after reconfguraton has been mplemented. METHODOLOGY OF IGA A conventonal GA seems to fnd global optmum when operatng on a large scale systems and t cannot mantan constant optmzaton response tme (Gumaraes et al., 2010; Rtthpakdee et al., 2013). Thus, mplementng IGA n fact solve the constrant encountered by conventonal genetc algorthm, unravel the global optmum faced n large scale systems and reduce the computatonal tme. It has three essental genetc operators whch are selecton, crossover and mutaton. The process of IGA appled n ths study s 1030

3 Fg. 2: Flowchart of IGA presented below n Fg. 2 whereby the mprovement s made towards the selecton operator. The IGA procedure was brefly explaned n accordance of ther genetc defnton followed by the workng prncple of the IGA. Intalzaton: The avalablty for number of swtches n the network reconfguraton s randomly generated from number of populaton selected for the ntal mplementaton of the algorthm. chromosomes. The ndvduals or chromosomes are evaluated accordng to ther ftness value (L et al., 2010; Shakeran et al., 2010). For IGA appled to dstrbuton network, the ndvduals or chromosomes s represented by number of swtches whle power losses for each swtches denotes ftness value for each ndvduals or chromosomes. The evaluaton process s employed whereby at ths stage ftness value s beng calculated wth the followng equaton: Ftness functon: Genetc Algorthm s about fndng the best generaton from generated ndvduals or 1031 P= n n 1 1 A ( P P + Q Q ) + B ( Q P PQ ) (1)

4 Fg. 3: Workng prncple of IGA R A = cos( δ δ ) V V (2) P loss (%) = P total N x 100 % where, P, Q = Real and reactve power at bus, respectvely P, Q = Real and reactve power at bus, respectvely R = Lne resstance of bus and V, V = Voltage magntude of bus and, respectvely δ, δ = Voltage angle of bus and, respectvely P loss = Power losses for each te swtches = Total power loss P total (3) Determnng the termnaton crtera: When applyng ths IGA, termnaton condton has to be specfed at the earler stage of mplementng the algorthm. The termnaton condton s shown as: (a) Number of populaton = 30 Number of ndvduals for each populaton = 5 Maxmum number of teratons = 10 An mproved genetc algorthm: Selecton process has been selected to be mproved n ths algorthm. The mproved process s prmarly depcted as shown n Fg. 3. Frstly, the frst selecton consstng of randomly lsted 30 populatons s done towards the network confguraton wthn the termnaton crtera specfed. Whle the termnaton crtera are beng acheved, the populaton obtaned wll experence the second process of selecton conssts of only 5 randomly lsted populatons and then the re-rankng process takes place to put the second selecton of populaton n ascendng 1032 (b) Fg. 4: DNA of genetc operators, (a) an example of crossover technque (b) an example of mutaton n one generaton technque order of power losses, so as to assst the researcher for fast and optmal soluton of total power losses obtaned for each te swtches. Crossover: Ths s the crtcal feature of IGA because t hugely accelerates search n the earler evoluton of a

5 populaton and also gudes an effectve combnaton of sub solutons between non-smlar chromosomes. The two offsprng are generated from two parents where each offsprng holds few genetc features of each parents. Hence, applyng ths same concept of conventonal genetc algorthm, crossover operator smply pcks any combnaton of 5 swtches under 2 nd selecton wth probabltes of 0.7 and the result s beng contnued to mutaton operator. The process contnues untl the maxmum number of teratons s acheved. Mutaton: Conventonal GA apples mutaton operator to each offsprng that beng produced to avod of trappng at local optmum and ths operator only creates small changes to the offsprng to stmulate dversty. Fgure 4 portrays the DNA of IGA whereby Fg. 4a and b are the examples of crossover and mutaton technques accordngly. CASE STUDY The ams of the proposed algorthm are to reduce total power losses and to mprove the voltage profle of the ntal radal 33-bus test system. In order to verfy the performance of network reconfguraton, 4 cases of crossover probablty (cp) and mutaton probablty (cm) are beng consdered, as lsted n the Table 1. It can be seen from Table 2, the four cases nvolvng the alteraton of crossover and mutaton probabltes essentally reduce the total power losses obtaned to the network reconfguraton. As for case 1, a total of 67.2 kw power losses are reduced whereas only 61.2 kw s reduced for case 2. In the case of cm adustment, the combnaton of te swtches 15, 27, 33, 8 and 12, respectvely n case 1 produces kw of total power losses meanwhle the same adustment n case 2 produces kw wth te swtches of 28, 4, 9, 14 and 16, respectvely. Hence, ths shows that low value of cm, 0.4 s able to produce greater reducton of power losses compared to hgh value of cm, 0.6. Meanwhle for case 3 and 4 wth the adustment of cp, case 4 demonstrates the hghest percentage of total power losses reducton whch s 21.10% whereas case 3 s only 16.44%. By havng kw of total power losses wth te swtches 13, 7, 15, 27 and 10, respectvely, case 4 s far better than case 3 that shows total power losses of kw wth te swtches Table 1: 4 Cases consdered Case 1 Case 2 Case 3 Case 4 Crossover probablty (cp) Mutaton probablty (cm) Table 2: Te swtches and power losses After reconfguraton Base case Case 1 Case 2 Case 3 Case 4 Te swtches 34, 35, 36, 37, 38 15, 27, 33, 8, 12 28, 4, 9, 14, 16 8, 6, 5, 9, 16 13, 7, 15, 27, 10 Total power losses (kw) Total power losses reducton (kw) Percentage of total power losses reducton (%) Fg. 5: Voltage profle for case 1 and

6 Fg. 6: Voltage profle for case 3 and 4 Table 3: Comparson of GA and IGA for case 4 After reconfguraton for case 4 Methods Te swtches Percentage of total power losses reducton (%) GA 9, 4, 31, 7, IGA 13, 7, 15, 27, , 6, 5, 9 and 16, respectvely. Nevertheless, t can be sad that a hgh cp value of 0.7 yelds a better reducton of power losses compared to the low value of cp, 0.5. As depcted n Fg. 5, voltage profles for case 1 and 2 show only a slght dfference at each number of buses. Intally, the voltages for both cases are at maxmum before they dropped slghtly from bus number 4 untl bus number 8, whereby the voltage for case 1 dropped more than that of case 2. For bus number 9 untl bus number 15, case 2 has greater fall compared to case 1, but reflectng the close proxmty of the respectve voltage profles for both cases. The voltage profle between case 1 and 2 starts to dffer extensvely from bus number 16 to bus number 18 and both cases acheve the maxmum value agan at bus number 19. From bus number 20 onwards, case 1 and 2 show small dssmlartes n ther voltage profles, but the dssmlartes start to ncrease from bus number 31 to bus number 33, wth the bus number 33 possesses the largest dfference. Fgure 6 dsplays the voltage profle obtaned for cases 3 and 4 after the alteraton of cp value takes place. From a bref overvew, case 3 shows vast dfference of voltage profle compared to that of case 4. At bus number 3, voltage profle for case 3 abruptly dropped to p.u whle case 4 mantaned ts voltage profle near to the maxmum value. However the pattern changed at bus number 9, where the value for case 4 excessvely falls to p.u. Bus number 19 shows great mprovement n voltage profle where t ncreases from p.u for case 4 and p.u for case 3 to the maxmum value. Then at bus number 20, the voltage profle n both cases drops whereby case 4 s at p.u and case 3 at p.u. For the bus number 21 and onwards, cases 3 and 4 happened to be regulated by ther prevous bus number value and contnues untl bus number 33. After analyzng all four cases for the regulated cp and cm values, the mnmum cm value n case 1 and maxmum cp value n case 4 exhbt better results compared wth the other two cases. Genetcally, the chosen bts n the next reproducton are complemented by mutaton and crossover s the possblty of producng a good offsprng n the next reproducton, thus showng that a hgh cp value and a low cm value s needed to yeld good offsprng and at the same tme mantanng ther genetc dversty. In ths study, case 4 s capable of demonstratng the hghest reducton n power losses and greatest mprovement of voltage profle. In order to verfy the results obtaned n case 4, t s compared wth the conventonal GA, as shown n Table 3, usng the same value of cp and cm. After the smulaton, t shows that the IGA has 21.10% reducton of power losses but the conventonal GA only shows reducton of 15.60%. The probabltes utlzed for both conventonal GA and IGA are smlar but dfferent set of te swtches are produced due to the algorthm mplementaton that randomly search the set of te swtches dentfcaton wth the best power losses reducton. As depcted n Table 3, conventonal GA fnds 9, 4, 31, 7 and 13 as ts best set of te swtches whle IGA search for the best set of te swtches from 13, 7, 15, 27 and 10, respectvely. Instead of usng conventonal GA, a better mprovement n voltage profles can also be attaned by usng the mproved GA.

7 CONCLUSION An adustment of crossover and mutaton probabltes proposed n ths study s used to mprove the performance of conventonal GA n terms of total power losses reducton and voltage profle mprovement. The proposed algorthm s appled to a radal 33-bus test system, smulated nto MATLAB software thus showng that case 4 produced the hghest reducton of power losses and the best mprovement of voltage profle compared to conventonal GA and the other 3 cases nvolved. However, for further valdaton of the proposed algorthm, the values of cp and cm can be vared n a wder range. ACKNOWLEDGMENT The authors would lke to thank Unversty Teknkal Malaysa Melaka (UTeM) for provdng fnancal and academc support. REFERENCES Cheragh, M. and P. Ramezanpour, An effcentfast method for determnng mnmum loss confguraton n radal dstrbuton networks based on senstvty analyss. Proceedng of IEEE Internatonal Power Engneerng and Internatonal Conference (PEDCO, 2012). Malacca, Malaysa, pp: Gumaraes, M.A.N., C.A. Castro and R. Romero, Dstrbuton systems operaton optmsaton through reconfguraton and capactor allocaton by a dedcated genetc algorthm. IET Gener. Transm. Ds., 4(11): Hu, Y., N. Hua, C. Wang, J. Gong and X. L, Research on dstrbuton network reconfguraton. Proceedng of Internatonal Conference on Computer, Mechatroncs, Control and Electronc Engneerng (CMCE, 2010), pp: L, D.D., C. He and H.Y. Shu, Optmzaton of electrc dstrbuton system of large offshore wnd farm wth mproved genetc algorthm. IEEE T. Rehabl. Eng., pp: 1-6. Nza Samsudn, S., Electrcty breakdown management n Malaysa: A case study on Tenaga Nasonal Berhad. M.A. Thess, Insttute of Technology Management and Enterpreneurshp, UTeM, pp: Rtthpakdee, A., A. Thammano and N. Premasathan, A new selecton operator to mprove the performance of genetc algorthm for optmzaton problems. Proceedng of IEEE Internatonal Conference on Mechatroncs and Automaton (ICMA, 2013). Takamatsu, Japan, pp: Shakeran, R., H. Tavakkol, S.H. Kamal and M. Hdayat, Improved genetc algorthm for loss and smultaneously relablty optmzaton n radal dstrbuton systems. Proceedng of 3rd Internatonal Conference on Advanced Computer Theory and Engneerng, 4: Shamsudn, N.H., M.S. Momat, A.F.A. Kadr, M.F. Sulama and M. Sulaman An optmal dstrbuton network reconfguraton and szng of dstrbuted generaton usng modfed genetc Algorthm. Int. J. Appl. Eng. Res., 9(20): Sulama, M.F., H. Mokhls and H.I. Jaafar, A DNR usng evolutonary PSO for power loss reducton. J. Telecommun. Electron. Comput. Eng., 5(1). Sulama, M.F., N.H. Shamsudn, H.I. Jaafar, W.M. Dahalan and H. Mokhls, 2014a. A DNR and DG szng smultaneously by usng EPSO. Proceedng of 5th Internatonal Conference on Intellgent Systems Modellng and Smulaton, pp: Sulama, M.F., M.S. Shdan, W.M. Dahalan, H. Mokhls, M.F. Baharom and H.I. Jaafar, 2014b. A 16kV dstrbuton network reconfguraton by usng evolutonarng programmng for loss mnmzng. Int. J. Appl. Eng. Res., 9(10): Zhao, F., L. Ge and W. L, Applcaton of antgenetc algorthm n reactve power optmzaton of dstrbuton network. Proceedng of Asa-Pacfc Power and Energy Engneerng Conference (APPEEC, 2012), pp: 1-4. Zhao, H., Y. Xe, N. Zheng and G. Wang, Improved genetc algorthm for reactve power optmzaton of dstrbuton system. Proceedng of 6th Internatonal Conference on Advances n Power System Control, Operaton and Management, pp: