J. Electrical Systems 12-4 (2016): Regular paper. Fault Identification in an Unbalanced Distribution System Using Support Vector Machine

Transcription

1 Soph Shlpa Gururaapathy 1 Hazle Mokhls 1* HazleeAzl Illas 1 AbHalm Abu Bakar 2 LlkJamlatul Awaln 3 J. Electrcal Systems 12-4 (216): Regular paper Fault Identcaton n an Unbalanced Dstrbuton System Usng Support ector Machne Fast and eectve ault locaton n dstrbuton system s mportant to mprove the power system relablty. Most o the researches rarely menton about eectve ault locaton consstng o aulted phase ault type aulty secton and ault dstance dentcaton. Ths work presents a method usng support vector machne to denty the aulted phase ault type aulty secton and dstance at the same tme. Support vector classcaton and regresson analyss are perormed to locate ault. The method uses the voltage sag data durng ault condton measured at the prmary substaton. The aulted phase and the ault type are dented usng three-dmensonal support vector classcaton. The possble aulty sectons are dented by matchng voltage sag at ault condton to the voltage sag n database and the possble sectons are ranked usng shortest dstance prncple. The ault dstance or the possble aulty sectons sthen dented usng support vector regresson analyss. The perormance o the proposed method was tested on an unbalanced dstrbuton system rom SaskPower Canada. The results show that the accuracy o the proposed method s satsactory. Keywords: Support ector Machne; Faulted Phase; Fault type; Faulty secton; Fault dstance. Artcle hstory: Receved 8 June 216 Accepted 25 October Introducton Dstrbuton systems supply electrc power to customers and occupy an mportant role n power system. A survey n [1] shows that more than 8% o the nterrupton n dstrbuton systems was caused by aults whch damaged the equpment and led to power outage to every customer on the system. Ths stuaton has orced electrcal power utltes to provde hgh relable and qualty power supply [2]. Hence to mantan contnuous power supply to customers aulty lne has to be dented and solated rom the system. An eectve ault locaton dentcaton n dstrbuton systems should be able to denty the aulted phase ault type aulty secton and ault dstance. Identcaton o aulted phase ault type and aulty secton urther helps to know the most requent type o ault occurred n partcular system. Thusproper mantenance can be taken to mnmze ts occurrence n the uture. arous knowledge based algorthms have been used to denty ault such as the Artcal Neural Network (ANN) [3] Wavelet Transorm (WT) Fuzzy Theory Matchng approach and Support ector Machne (SM). The method usng voltage sag characterstc [4 5] dentes the ault type by comparng the pattern o pre-ault voltage wth the voltage durng ault. However or ault ar rom the measurement locaton the derent s not * Correspondng author: H. Mokhls Department o Electrcal Engneerng Faculty o Engneerng Unversty o Malaya 563 Kuala Lumpur Malaysa E-mal: hazl@um.edu.my 1 Department o Electrcal Engneerng Faculty o Engneerng Unversty o Malaya 563 Kuala Lumpur Malaysa 2 UM Power Energy Dedcated Advanced Centre (UMPEDAC) Level 4 Wsma R&D Unversty o Malaya JalanPantaBaru 5999 Kuala Lumpur Malaysa 3 Unversty Kuala Lumpur Electrcal Engneerng Secton Internatonal College Brtsh Malaysan Insttute Batu 8 Jalan Sunga Pusu 531Gombak Selangor Malaysa Copyrght JES 216 on-lne : ournal/esrgroups.org/es

2 J. Electrcal Systems 12-4 (216): notceable and may lead to wrong dentcaton o ault type. Fuzzy set wasproposed n [6-8] and neural network n [9] or ault type classcaton.the methods such as n [1 11] denty ault type or transmsson systems usng SM. The method n [1] uses zero sequence and three phase currents to denty the aulted phase n transmsson system. In [11] ault classcaton usng prncpal component analyss and SM s proposed. SM wasused or aultclasscaton and sectondentcaton n [12]. ANN wasproposed n [13] whch uses voltage and current to classy Double Lne to Ground Fault. The method n [13]dentedthe ault type and aulty secton. A combnaton o SM and Wavelet transorm was proposed n [14] or predcton o ault type and locaton. It uses voltage and current sgnals to locate ault n transmsson system. Fault locaton wth WT and wavelet packet transorm (WPT) combnng artcal neural network was proposed n [15]. A hybrd approach usng wavelet transorm and SM was suggested n [16] or precse ault locaton n transmsson lnes. A method to denty aulty secton and ault dstance by usng matchng approach was proposed n [17 18]. The method n [18] also ranks the possble secton based on prorty. The advantage o the method s that t can be used or any number o measurements n the network. SM was used n [19] to denty the ault type aulty secton and dstance together o aulted lnes n a transmsson system. Most o the prevous ault locaton methods dagnose aults or transmsson systems but not n unbalanced dstrbuton systems. Derent rom transmsson systems dstrbuton systems have more complex topologcal structures wth multple laterals. The methods ocus on ndng the ault type or the ault dstance separately. None o the researches presents eectve ault locaton dentcaton by consderng the aulted phase ault type aulty secton and ault dstance at the same tme n an unbalanced dstrbuton system. Consderng ths lmtaton ths work tends to denty aulted phase ault type aulty secton and ault dstance n a sngle method. The ault type tsel ncludes the aulted phase whch are Sngle Lne to Ground Fault at phase a (SLGF a ) Sngle Lne to Ground Fault at phase b (SLGF b ) Sngle Lne to Ground Fault at phase c (SLGF c ) Lne to Lne Fault at phase ab (LLF ab ) Lne to Lne Fault at phase bc (LLF bc ) Lne to Lne Fault at phase ca (LLF ca ) Double Lne to Ground Fault at phase ab (DLGF ab ) Double Lne to Ground Fault at phase bc (DLGF bc ) Double Lne to Ground Fault at phase ca (DLGF ca ) and Three phase to ground ault at phase abc (LLLGF abc ). The proposed method locates ault usng three-dmensonal (3D) analyss o SM. The aulted phase and the ault type are dented usng multclass Support ector Classcaton (SC). The aulty secton s dented by estmatng ault resstance usng Support ector Regresson (SR). The possble sectons are dented by matchng the actual voltage sag wth the smulated data and the most possble sectons are ranked. Fnally the ault dstance s estmated usng SR analyss. Ths manuscrpt s organzed n 5 sectons. Secton 2 descrbes the proposed methodology o the work. Secton 3 presents the test system and secton 4 gves the test results o the proposed method. Secton 5 concludes the ndng o ths work. 787

3 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM 2. Proposed Methodology The proposed method utlzes voltage sag magntude or 3 phases (phase a b and c) to denty the ault n unbalanced dstrbuton systems. It rst dentes the ault type ncludng aulted phase then the aulty secton and nally the ault dstance. The llustraton o proposed method s shown n Fgure 1. ( a b c ) 2.1. Database establshment Fgure 1: Illustraton o the proposed method The proposed method s mplemented usng SM whch requres a tranng set o voltage sag data or processng. The database establshment s llustrated n Fgure 2. The steps nvolve are as ollows: 1. Sngle Lne to Ground Fault at phase a (SLGF a ) s smulated at all nodes o dstrbuton system wth Ω resstance 2. oltage sag magntude at phase a b and c are recorded rom the measurement node 3. The smulaton s repeated or ault resstance o 2Ω 4Ω and 6Ω resstance 4. Steps 1 to 3 are repeated or other ault types o SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc. R (x) R ( x+ R ( x+2) R ( x+3) Fgure 2: Tranng data establshment 788

4 J. Electrcal Systems 12-4 (216): Fault type dentcaton The type o ault can be dented usng one versus all concept o multclass SC. The proposed method uses voltage sag data at ault ( a b and c ) as nput or SC. The desred output s the type o the ault. Fgure 3 descrbes the ault type classcaton usng SC. At each SC there are two possble nputs or classcaton class 1 and class. At rst the voltage sag data o SLGF a s consdered as class 1 and the remanng (SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc ) are consdered as class. SC nds the optmal hyper-plane between the two classes and dentes whether the nput data alls n class 1 or class. I the ault type s dented under class 1 then the ault type s nalzed as SLGF a. I the ault type s dented under class then a second step o classcaton takes place by consderng SLGF b as class 1 and the remanng (SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc ) as class. The process s contnued untl the actual ault type s dented. a b c 2.3. Faulty secton dentcaton Fgure 3: Fault type dentcaton usng SC Once the ault type s dented the aulty secton s dented. Faulty secton dentcaton conssts o ault resstance estmaton selecton o possble sectons and rankng analyss Fault resstance estmaton Fault resstance s estmated usng SR analyss. The voltage sag data rom database s traned usng radal bass uncton n SR. The voltage sag at ault condtons ( a b and est c ) are assgned as the nput to SR. The correspondng output ( R ) s the estmated ault resstance. The llustraton o the ault resstance estmaton s depcted n Fgure 4. a b est R c Fgure 4: Fault resstance estmaton usng SR 789

5 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM Selecton o possble sectons Once the ault resstance s estmated the possble aulty sectons are dented by comparng the voltage sag magntude n database wth the actual voltage sag magntude [17 2]. The ault resstance rom the database are selected such that est R < R < R ( x + ( where o R (x) and R ( x + are the resstance rom database or whch the voltage sag s smulated. The voltage sag data or each secton rom R (x) and R ( x + are analysed ndvdually. For example a aulty secton s between nodes and are consdered. The voltage sag magntudes wth a ault resstance between R (x) and ( x + are shown n Table 1. Node Table 1: oltage sag data or secton dentcaton Node R ( x ) R ( x ) R ( x ) R Fault resstance n database R ( x ) ( x + a b c R ( x ) R ( x ) R ( x ) Node R R ( x+ R ( x+ R ( x+ a b c R ( x+ R ( x+ R ( x+ a b c a b c Fgure 5 showsthe search boundary o secton s at resstance values R (x) and R ( x + n 3D. ) ( a b c correspondngto the measured voltage sag magntude at phase a b and c at ault condton. It can be seen that the measured voltage sag s not wthn the search boundary. To address ths problem the mnmum and maxmum voltage sag proles o two adacent ault resstances are consdered. b R ( x + R ( x+ R ( x+ R ( x+ a b c R ( x+ R ( x+ R ( x+ a b c ( a b c ) a b c R ( x ) a b c R (x) a c Fgure 5: oltage sag prole varaton or secton s and two derent resstances 79

6 J. Electrcal Systems 12-4 (216): The mnmum and maxmum voltage sags o the secton s are noted down. I the voltage sag at ault les between mnmum and maxmum o the secton rom database the correspondng secton s chosen as the aulty secton [17]. R a R ( x+ a (2) R b a R ( x+ b (3) R c b R ( x+ c (4) c Rankng Analyss Multple aulty sectons are possble n dstrbuton system due to the presence o lateral branches and sub-branches. Hence the most possble aulty secton s ranked usng the smlar concept o shortest dstance prncple [18]. The shortest dstance d s calculated between the ault pont and the lnear lne onng voltage sag rom database. The aulty secton whch yelds the shortest dstance among all possble aulty secton has a hgh prorty o the most possble aulty secton. Fgure 6 shows two possble aulty sectons s (nodes -) and m(nodes p-q). a b R c R ( x+ R ( x+ R ( x+ and represent a b the mnmum and maxmum values o voltage sag at secton s. R ( x+ R ( x+ R ( x+ a q b q c q s c a p b p c p represent the mnmum and maxmum values o voltage sag at secton m. ( a b c ) represents the voltage sag data dented durng the ault. d and s1 d represent the shortest dstance between the ault pont and the lne onng mnmum and s2 maxmum values o sectons s and m. a b c b d s1 R ( x+ R ( x+ R ( x+ a b c and R R a p b p R R a b R c p R c d s2 R ( x+ R ( x+ R ( x+ a q b q c q a c Fgure 6: Rankng analyss The shortest dstance d s1 or a secton s s calculated usng d s M M x + M y + M z = = (5) S S + S + S 2 x 2 y 2 z 791

7 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM where S { S S S } = s the drectng vector o lne onng R ( x+ R ( x+ R ( x+ a b c { N N N } x y x z y z N = s the drectng vector o lne onng ( a b c ) { M M M } M = s the cross product o vectors N and s. x y z a b a b R c R c and and 2.4. Fault Dstance estmaton Fault dstance s dented usng SR analyss. For the possble aulty sectons the voltage sag data at nodes rom the database (Table s traned usng SR to estmate the ault dstance. The tranng nput and output or secton s s shown n Table 2. Here l represents the length o the lne secton. Table 2: Tranng data or ault dstance calculaton Input tranng data Output tranng data R ( x ) R ( x ) R ( x ) a b c R ( x+ R ( x+ R ( x+ a b c R ( x ) R ( x ) a b R ( x ) c l R ( x+ R ( x+ R ( x+ a b c l The llustraton or ault dstance estmaton s shown n Fgure 7. The voltage sag data durng the ault ) and the estmated ault resstance ( a b c nput to SR. The correspondng output s the ault dstance d. est R are assgned as a b c est R d Fgure 7: Fault dstance estmaton 3. Test System The SaskPower network s a radal dstrbuton network consstng o unbalanced lnes and unbalanced loads. The schematc dagram o test dstrbuton system s shown n Fgure 8. The system conssts o a 25k equvalent source sngle phase laterals three phase 792

8 J. Electrcal Systems 12-4 (216): laterals and 2 lne sectons made up o derent conductor. A node number s ndcated along the lne o the test system. The parameters o equvalent source lne data and the load data can be obtaned rom [ ]. The dstrbuton system s modelled usng PSCAD power system smulaton sotware. The cables are modelled as constant mpedance load usng PI model. The voltage sag data s recorded n measurement node nearer to node 1 o dstrbuton system. The voltage sag database s created by smulatng ault at all nodes o the dstrbuton system. The perormance o algorthm s tested or varous ault types such as SLGF a SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc and or varous resstances o Ω 1Ω 3Ω and 5Ω. The measured voltage sag data at the ault s analysed usng a MATLAB programmng code. 4. Test Results Fgure 8: Schematc dagram o SaskPower Dstrbuton system For tranng purpose smulatons were perormed or ault at the nodes o the dstrbuton system at Ω 2Ω 4Ω and 6Ω resstance. A total o 84*3 voltage sag data are utlzed or tranng usng SM. For testng purpose aults at the mddle o the lne secton at Ω 1Ω 3Ω and 5Ω resstance Fault type classcaton The proposed method s tested or ten ault types (SLGF a SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc ). The subscrpt n the ault type represents the aulted phase. The ault type and the aulted phase are dented usng 3D multclass SC. SC s traned wth 84 voltage samples usng radal bass uncton (RBF) or classcaton o 1 output. The support vectors dented usng SC durng ault type classcaton are tabulated n Table

9 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM Table 3: Support vectors o SC Types o ault Class 1 Class Idented Support vectors SLGF a SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc The hyper plane obtaned or ault type classcaton s shown n Fgure 9(a) to Fgure 9(). Fgure 9(a) gves the total o 84 voltage sag data and 3D non-lnear hyperplane dented or SLGF a. Class 1 represents the data o SLGF a and the remanng (SLGF b SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc ) as Class. The x- axs represents the voltage at phase a y-axs represents the voltage at phase b and z-axs represents the voltage at phase c. I the ault type s not SLGF a then urther classcaton s carred out by consderng SLGF b as class 1 as gven n Fgure 9(b). Smlarly the hyperplane or classcaton o ault types SLGF c LLF ab LLF bc LLF ca DLGF ab DLGF bc DLGF ca and LLLGF abc are shown n Fgure 9(c) to Fgure 9() 9(a) 3D hyper plane or SLGF a 9(b) 3D hyper plane or SLGF b 794

10 J. Electrcal Systems 12-4 (216): (c) 3D hyper plane or SLGF c 9(d) 3D hyper plane or LLF ab 9(e) 3D hyper plane or LLF bc 9() 3D hyper plane or LLF ca 9(g) 3D hyper plane or DLGF ab 9(h) 3D hyper plane or DLGF bc 795

11 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM 4.2. Faulty secton 9() 3D hyper plane or DLGF ca and LLLGF abc Fgure 9: 3D classcaton usng SC The test results o possble aulty sectons and rank o the correct secton at Ω resstance s llustrated n Table 4. For analyss the test sectons 1-2 and 7-8 nodes (Man at eeder) nodes (Branch at eeder) 18-2 nodes (Sub branch at eeder) are consdered. From the results t can be noted that sngle aulty sectons were selected or test sectons 1-2 and 7-8 or all types o ault. Ths s because rom node 1 to node 2 and node 7 to node 8 s completely a radal lne and there are no parallel lne sectons. Also the rank dented or the ault type o SLGF a SLGF b and SLGF c ; LLF ab LLF bc and LLF ca ; DLGF ab DLGF bc and DLGF ca are the same. Ths s due to only the aulted phase s nterchanged. Secton Number Table 4: Faulty sectons and rank o the correct secton atω resstance Test Secton Selected Faulty Secton Rank Number o the Actual Faulty Secton SLGF a / SLGF b / SLGF c LLF ab / LLF bc / LLF ca DLGF ab / DLGF bc / DLGF ca LLLGF abc The test results o aulty secton are also analysed or varous other ault resstances o 1Ω 3Ω and 5Ω. The calculated ault resstance and the rankng usng shortest dstance prncple or test secton and 18-2 nodes are tabulated n Table 5. It can be noted that the calculated ault resstance s closer to the actual resstance. 796

12 J. Electrcal Systems 12-4 (216): Table 5: Fault resstance and rankng o the aulty sectons SLGF a SLGF b LLF ab LLF bc DLGF ab DLGF bc LLLGF Faul SLGF Actual c LLF ca DLGF abc ca ty Calculate Calculat ault Calculated Calculated sect d ed resstance resstance Rank Rank resstance Rank Rank on resstance resstanc (Ω) (Ω) (Ω) e (Ω) The overall rankng perormance o the proposed method s shown n Fgure 1. The x- axs represents the rank and y-axs represents the number o possble canddate dented n the rankng. The test cases are repeated or resstances o Ω 1Ω 3Ω and 5Ω or ault at the mdpont o all 2 lne sectons. It shows that most o the possble aulty sectons are ound correctly at the rst and second ranks or md-pont tests at all test sectons. For Ω resstance 14 aulty sectons are correctly dented n the rst rank or LLF 15 sectons or SLGF and DLGF and 18 sectons or LLLGF. The aulty secton perormances o LLF (2 sectons) and DLGF (1 secton) have rankngs up to the thrd rank. For other ault resstances o 1Ω 3Ω and 5Ω the result shows that all o the sectons can be determned wthn rst sx ranks No. o possble canddate Ω 1Ω 3Ω 5Ω No. o possble canddate Ω 1Ω 3Ω 5Ω Rank 1 Rank 2 Rank 3 Rank 4 Rank 5 Rank 1Rank 2Rank 3Rank 4Rank 5Rank 6 1(a) SLGF a SLGF b SLGF c 1(b) LLF ab LLF bc LLF ca 797

13 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM No. o possble canddate Ω 1Ω 3Ω 5Ω No. o possble canddate Ω 1Ω 3Ω 5Ω 2 2 Rank 1Rank 2Rank 3Rank 4Rank 5Rank 6 Rank 1Rank 2Rank 3Rank 4Rank 5Rank 6 1(c) DLGF ab DLGF bc DLGF ca 1(d) LLLGF abc Fgure 1: Overall rankng perormance 4.3. Fault dstance calculaton The ault dstance sanalysed usng SaskPower dstrbuton network or ault at the mdpont o all lne secton. Fgure 11 shows the percentage error o calculated ault dstance or SLGFat resstances o Ω 1Ω 3Ω and 5Ω. The test results o ault dstance or SLGF a SLGF b and SLGF c are the same because the voltage sag at phase a phase b and phase c are ust nterchanged. A maxmum percentageerror o 24.7% sobtaned n SLGF (secton 1-2)at a aultresstance o 1Ω. Percentage error (%) Ω 1 Ω 3 Ω 5 Ω Faulty secton Fgure 11: Calculated ault dstance or SLGF a / SLGF b / SLGF c The percentageerror or LLF ab LLF bc and LLF ca at resstances o Ω 1Ω 3Ω and 5Ωsshown n Fgure 12. A maxmum percentageerror o 11.3% sobtaned n test secton 1-2 at1ωresstance. All other test sectons havelowerpercentageerror. Percentage error (%) Fgure 12: Calculated ault dstance or LLF ab / LLF bc / LLF ca Faulty secton Ω 1 Ω 3 Ω 5 Ω

14 J. Electrcal Systems 12-4 (216): The percentageerror o ault dstance or DLGF ab DLGF bc and DLGF ca at resstances o Ω 1Ω 3Ω and 5Ωsshown n Fgure 13. A maxmum o 23% sdented n test secton 1-2 at 1Ω resstance. Percentage error (%) Ω 1 Ω 3 Ω 5 Ω Faulty secton Fgure 13: Calculated ault dstance or DLGF ab / DLGF bc / DLGF ca Fgure 14 gves the percentage error o LLLGF abc.a maxmum percentage error o 3% s obtaned at 1 Ω resstance (at secton 1-2) or LLLGF abc. In ths the devaton rom the actual ault dstance s 362 meters whch s a small dstance compared to the whole dstrbuton system. The percentage error o ault dstance at other resstance o Ω 3 Ω and 5 Ω are less than 3% error. Thereore the proposed method has managed to denty the ault dstance wth greater accuracy. Percentage error (%) Ω 1 Ω 3 Ω 5 Ω Faulty secton Fgure 14: Calculated ault dstance or LLLGF abc 5. Conclusons An approach usng three-dmensonal support vector classcaton and regresson analyss or locatng ault has been successully proposed n ths work. The ault type and the aulted phase are dented usng SC. The method classes all 1 types o aults by dentyng the hyper plane between classes. The aulty secton was dented by usng matchng approach and rankng the most possble aulty secton. The possble aulty secton was ranked usng three-dmensonal shortest dstance prncple. The proposed work shows that the aulty sectons were dented wthn rst sx rankng and all o the aulty sectons can be ranked. Also ault dstances or the possble aulty sectons were dented usng SR analyss. A maxmum error o 3% was obtaned n the test cases. Thereore the proposed method has the potental to be used to denty the aulted phase ault type aulty secton and ault dstance or varous ault resstances. 799

15 S. Shlpa Gururaapathy et al: Fault Identcaton n an Unbalanced Dstrbuton System Usng SM Acknowledgement The authors thank the Malaysan Mnstry o Educaton and Unversty o Malaya or supportng ths work through research grant o HIR (H-161-D48) and FRGS (FP26-212A). Reerences [1] U.-C. P. S. O. T. Force Abraham S. et al. US-Canada Power System Outage Task Force Fnal Report on the August Blackout n the Unted States and Canada: Causes and Recommendatons 24. [2] Y. Menchaou M. Zahr M. Habb and H. E. Markh Extenson o the Accurate oltage-sag Fault Locaton Method n Electrcal Power Dstrbuton SystemsJournal o Electrcal Systems [3] M. R. Garous M. R. Shakaram and F. Namdar Detecton and classcaton o power qualty dsturbances usng parallel neural networks based on dscrete wavelet transorm Journal o Electrcal Systems [4] C. T. Suresh Kamble oltage Sag Characterzaton n a Dstrbuton Systems: A Case Study Journal o Power and Energy Engneerng [5] D. M. K. Namrata B. Pawar Generaton o Derent Types o oltage Sag Usng Matlab/Smulnk Internatonal Journal o Engneerng and Innovatve Technology [6] A. K. Pradhan A. Routray and B. Bswal Hgher order statstcs-uzzy ntegrated scheme or ault classcaton o a seres-compensated transmsson lneieee Transactons on Power Delvery [7] B. Das and J.. Reddy Fuzzy-logc-based ault classcaton scheme or dgtal dstance protectonieee Transactons on Power Delvery [8] B. Das Fuzzy logc-based ault-type dentcaton n unbalanced radal power dstrbuton systemieee Transactons on Power Delvery [9] A. Flores E. Qules E. García and F. Morant Novel Formulaton usng Artcal Neural Networks or Fault Dagnoss n Electrc Power Systems A Modular ApproachJournal o Electrcal Systems [1] U. B. Parkh B. Das and R. Maheshwar Fault classcaton technque or seres compensated transmsson lne usng support vector machne Internatonal Journal o Electrcal Power & Energy Systems 32(7) [11] Y. Guo K. L and X. Lu Fault Dagnoss or Power System Transmsson Lne Based on PCA and SMsIntellgent Computng or Sustanable Energy and Envronment [12] P. K. Dash S. R. Samantaray and G. Panda Fault Classcaton and Secton Identcaton o an Advanced Seres-Compensated Transmsson Lne Usng Support ector Machne IEEE Transactons on Power Delvery [13] A. J. Prarthana Warlyan A.S.Thoke and R.N.Patel Fault Classcaton and Faulty Secton Identcaton n Teed Transmsson Crcuts Usng ANNInternatonal Journal o Computer and Electrcal Engneerng [14] S. Ekc Support ector Machnes or classcaton and locatng aults on transmsson lnesappled Sot Computng 12(6) [15] P. Ray B. K. Pangrah and N. Senroy Hybrd methodology or ault dstance estmaton n seres compensated transmsson lneiet Generaton Transmsson & Dstrbuton [16] A. Saber A. Emam and R. Amer Dscrete wavelet transorm and support vector machne-based parallel transmsson lne aults classcaton IEEJ Transactons on Electrcal and Electronc Engneerng 215. [17] H. Mokhls and H. L Non-lnear representaton o voltage sag proles or ault locaton n dstrbuton networkselectrcal Power and Energy Systems [18] L. J. Awaln H. Mokhls A. Abu Bakar H. Mohamad and H. A. Illas A generalzed ault locaton method based on voltage sags or dstrbuton networkieej Transactons on Electrcal and Electronc Engneerng 8 S38-S [19] B. Ravkumar D. Thukaram and H. P. Khncha Applcaton o support vector machnes or ault dagnoss n power transmsson systemiet Generaton Transmsson & Dstrbuton [2] L. H. Mokhls H Khald AR. The applcaton o voltage sags pattern to locate a aulted secton n dstrbuton network IEEE Internatonal Revew o Electrcal Engneerng [21] J. Mora-Flòrez J. Meléndez and G. Carrllo-Cacedo Comparson o mpedance based ault locaton methods or power dstrbuton systems Electrc Power Systems Research 78(4) [22] L. Seung-Jae C. Myeon-Song K. Sang-Hee J. Bo-Gun L. Duck-Su A. Bok-Shn et al. An ntellgent and ecent ault locaton and dagnoss scheme or radal dstrbuton systems IEEE Transactons on Power Delvery