您好,欢迎访问三七文档
当前位置:首页 > 商业/管理/HR > 经营企划 > 变压器绝缘故障类型的改进型RBF神经网络识别算法李浩
5DOI:10.13234/j.issn.2095-2805.2018.5.167:TM41:ARBF,,(,454000):,8,RBF(radialbasisfunction)。RBF、,。:,90%。:;;RBF;;IdentificationAlgorithmforTransformerInsulationFaultTypesBasedonImprovedRBFNeuralNetworkLIHao,WANGFuzhong,WANGRui(SchoolofElectricalEngineeringandAutomation,HenanPolytechnicUniversity,Jiaozuo454000,China)Abstract:Toaccuratelydiagnosetheinternallatentfaulttypesofapowertransformer,anovelradialbasisfunction(RBF)neuralnetworkalgorithmisproposedbyanalyzingthegasproductionundereightlatentinternalinsulationfaulttypes,suchasoiloverheatingandpartialdischarginginoilpaperinsulation.Thisalgorithmisimprovedbyartificialimmunenetworkalgorithmandparticleswarmoptimizationalgorithm.ThispaperfocusesonthecompositionprincipleoftransformerfaultdiagnosismodelbasedonRBFneuralnetwork,themethodfordeterminingthecenterofhiddenlayerinthefaultmodelbasedonartificialimmunenetworkalgorithm,andthemethodofnetworkweightoptimizationbasedonparticleswarmoptimizationalgorithm.Simulationexperimentsarecarriedout,showingthattheproposedalgorithmcaneffectivelyidentifytheinsulationfaulttypesatanaccuracyofhigherthan90%.Keywords:powertransformer;faultdiagnosis;RBFneuralnetwork;artificialimmunenetwork;particleswarmoptimizationalgorithm,,[1,2]。,。DGA(dis-solvedgasanalysis),,[3]。,。,,[4-9]、[10]、[11-12]。RBF(radialbasisfunction):2016-08-07;:2018-01-11:(132107000027)ProjectSupportedbyFoundationfortheIntegrationofIndustry,EducationandResearchofHenanProvince(132107000027)JournalofPowerSupplyVol.16No.5Sept.201816520189791Tab.1GasproductiontypesunderdifferenttypesoffaultCH4、C2H4H2、C2H6CH4、C2H4、CO、CO2H2、C2H6H2-CO、CO2-H2、CH4、COC2H2、C2H6、CO2H2、C2H2-H2、C2H2CH4、C2H4、CO2H2、C2H2、CO、CO2CH4、C2H4、C2H6、,、。、RBF[13]。[14]RBF,、,,;[15],RBF,RBF,。RBF,,、。,,。1:(、)()。,、,1。1,,4,8。(H2)、(CH4)、(C2H6)、(C2H4)、(C2H2)、(CO)(CO2)。,187,,,x1~x77,y1~y818。2RBF2.1RBF,[16]。,。,,、,RBF1。、。,,,7;2,;,,8。,sj(x)=准j(||x-cj||)j=1,2,…,a(1):sjj;准j(·)1685,:RBF;cj;||x-cj||xcj;a。、,准j(||x-cj||)=exp-(x-cj)T(x-cj)δ2jjj(2),,yk=aj=1Σωjksjk=1,2,…,m(3):ykk;m。2.2NX=[x1,x2,…,xN],xi=[xi1,xi2,xi3,xi4,xi5,xi6,xi7]T,Y=[yi1,yi2,yi3,yi4,yi5,yi6,yi7,yi8]T。,yk=ωTkΦ=aj=1Σωkjexp-||x-cj||2δ2jjjk=1,2,…,8(4):ωkk,ωk=[wk1,wk2,…,wka]T;Φ,Φ=[准1,准2,…,准a]T。(4),RBF。。2.3,RBF,,,。,,,[17-19]。,,。(1)。7xi=[xi1,xi2,xi3,xi4,xi5,xi6,xi7],xi1~xi77。(2)。[20,21],xi=[xi1,xi2,xi3,xi4,xi5,xi6,xi7],x'i=[x'i1,x'i2,x'i3,x'i4,x'i5,x'i6,x'i7],x'ip=xip7p=1Σxipp=1,2,…,7(5)(3)。,。,、cv,cv0.01,,。nd(ΣA,ΣB)=ni=1Σ(a[i]-b[i])2姨(6)(4)。NAb;(5)。x'iAbfi,fij=11+||x'i-xj||xj∈Ab(7)(6)。fi,Abn1RBFFig.1FaultdiagnosismodelofRBFneuralnetworky1y2y3y4y5y6y7y8ΣΣΣΣΣΣΣΣH2CH4C2H6C2H4C2H2COCO2ωijS1S2SjSaΦ1Φ2ΦjΦa……16979Ab{n};(7)。Ab{n},NcNc=ni=1Σround(Kscalefij)(8):round();Kscale。Ci。(8)。CiC*i,c*j=cj-α(cj-x'i)(9),α,,α。(9)。x'iC*i,f*iC*iξ%Ab*,Ab*σd;(10)。Ab*,σs,Ab*;(11)。。2.4RBFPSO(particleswarmoptimization)、[22-23]。GA(geneticalgorithm),、,,PSO。PSOP-V,。mR,,Pi=(Pi1,Pi2,…,PiR)、Vi=(Vi1,Vi2,…,ViR)。BestPi,BestGi,Vk+1id=ωVkid+C1R1(BestPid-Pkid)+C2R2(BestGid-Pkid)(10)Pk+1id=Pkid+Vk+1id(11):i=1,2,…,m;d=1,2,…,R;k;ω;BestPid、BestGidik;C1、C2;R1、R2[0,1]。PSO,,RBF。PSO,,。。1。C1、C2kmax,m,PiVi;2。RBF,BestPBestG。Fi=1NNi=1Σ(Y'i-Yi)2(12):Y'i;Yi;N。3BestP,BestP;BestG,BestG;4(10)(11),;5kmax,,;2。33.12013—2014220kV、150MV·A,300,120,180。3.217052Tab.2Diagnosticresult/%y1y2yy4y5y6y7y8y125232292y234430488y3880100y412120100y522202291y617170100y721183386y8412237488,:cv=0.01,N=20,n=6,Kscale=10,ξ%=10%,δd=0.5,δs=0.15。11。RBF,PSO,:m=40;C1=C2=2;ωmax=1.4;ωmin=0.4;kmax=500,810000000~00000001。0~1,1,2。,180,0.02,。3.3180RBF,2。,BP、RBF,3。2,,y3、y4y6100%,3,590%。23。4(1),、8。、、、。(2)RBF,,8,。,,90%。,:RBF2Fig.2Fitnesscurve3Tab.3Comparisonofdiagnosticaccuracyamongdifferentdiagnosticalgorithms/%BP18015787.4RBF18014178.318016692.20.0400.0380.0360.0340.0320.0300.0280.0260.0240.02220018016014012010080604020017179:[1],,,.[J].,2011,31(7):57-63.ZhengRuirui,ZhaoJiyin,ZhaoTingting,etal.Powertransformerfaultdiagnosisbasedongeneticsupportvectormachineandgrayartificialimmunealgorithm[J].Proceed-ingoftheCSEE,2011,31(7):57-63(inChinese).[2],,.RBF[J].(),2016,49(1):88-93.LiuJingyan,WangFuzhong,YangZhanshan.TransformerfaultdiagnosisbasedonRBFneuralnetworkandadaptivegeneticalgorithm[J].EngineeringJournalofWuhanUniver-sity,2016,49(1):88-93(inChinese).[3],,,.[J].,2007,41(6):722-726.WuXiaohui,LiuJiong,LiangYongchun,etal.Applicationofsupportvectormachineintransformerfaultdiagnosis[J].JournalofXi’anJiaotongUniversity,2007,41(6):722-726(inChinese).[4],,.BP[J].,2005,31(7):12-14.WangXuemei,LiWenshen,YanZhang.ApplicationstudyofBPnetworkusedinthefaultdiagnosisofpowertrans-former[J].HighVoltageEngineering,2005,31(7):12-14(inChinese).[5],,,.L-MBP[J].,2011,39(8):100-103.XiangWenqiang,ZhangHua,WangYuan,etal.Applica-tionofBPneuralnetworkwithL-Malgorithminpowertransformerfaultdiagnosis[J].PowerSystemProtectionandControl,2011,39(8):100-103(inChinese).[6],.LMBP[J].,2013,49(5):54-59.SongZhijie,WangJian.TransformerfaultdiagnosisbasedonBPneuralnetworkoptimizedbyfuzzyclusteringandLMalgorithm[J].HighVoltageApparatus,2013,49(5):54-59(inChinese).[7],.RBF[J].,2010,38(11):6-9.RenJing,HuangJiadong.
本文标题:变压器绝缘故障类型的改进型RBF神经网络识别算法李浩
链接地址:https://www.777doc.com/doc-6544927 .html