基于數(shù)據(jù)驅(qū)動(dòng)的配電網(wǎng)拓?fù)渥R(shí)別方法研究馬麗副研究員中國科學(xué)院電工研究所2024年9月21日西安團(tuán)隊(duì)介紹中國科學(xué)院電工研究所電網(wǎng)技術(shù)研究部，是國內(nèi)最早從事分布式能源和微網(wǎng)研究方向的團(tuán)隊(duì)之一，現(xiàn)任研究部主任為裴瑋研究員?，F(xiàn)有人員50余人，研究方向包括交直流配電網(wǎng)、新型儲(chǔ)能系統(tǒng)、新能源電力系統(tǒng)穩(wěn)定與控制、人工智能在電力系統(tǒng)中的應(yīng)用、分布式能源市場、綜合能源運(yùn)行優(yōu)化等。團(tuán)隊(duì)主持了科技部863項(xiàng)目、973課題、國家重點(diǎn)研發(fā)專項(xiàng)項(xiàng)目及課題10余項(xiàng)，自然科學(xué)基金重點(diǎn)及面上項(xiàng)目20余項(xiàng)，省部級(jí)及企業(yè)項(xiàng)目60余項(xiàng)，正在牽頭中國科學(xué)院先導(dǎo)B“新能源主導(dǎo)的電力系統(tǒng)穩(wěn)定控制新理論新技術(shù)”專項(xiàng)。牽頭獲得了北京市科技進(jìn)步一等獎(jiǎng)，中國仿真學(xué)會(huì)技術(shù)發(fā)明一等獎(jiǎng)、中國電力科學(xué)技術(shù)進(jìn)步一等獎(jiǎng)、電力科技創(chuàng)新一等獎(jiǎng)等獎(jiǎng)項(xiàng)。中國科學(xué)院電工研究所z■研究方向個(gè)人基本情況教育背景教育背景·2012-2017,博士，華北電力大學(xué)，電力系統(tǒng)及其自動(dòng)化(導(dǎo)師：張建華/劉念教授)·2008-2011,碩士，華北電力大學(xué)，電力系統(tǒng)及其自動(dòng)化(導(dǎo)師：張建華/劉念教授)·2004-2008,學(xué)士，華北電力大學(xué)，電氣工程及其自動(dòng)化工作經(jīng)歷工作經(jīng)歷求側(cè)管理、本地·2021.10-,副研究員，中國科學(xué)院電工研究所(裴瑋研究員團(tuán)隊(duì))·2020.9-2021.8,博士后，斯蒂文斯理工學(xué)院(合作導(dǎo)師：吳磊教授IEEEFellow)·2018.9-2020.8,博士后，威斯康辛大學(xué)密爾沃基分校(合作導(dǎo)師：王凌峰教授)·2011.4-2018.8,研發(fā)工程師，中國電力科學(xué)研究院配電研究所社會(huì)兼職社會(huì)兼職長時(shí)規(guī)模儲(chǔ)能重點(diǎn)實(shí)驗(yàn)室副主任EEESeniorMember;中國電機(jī)工程學(xué)會(huì)高級(jí)會(huì)員；中國電工技術(shù)學(xué)會(huì)會(huì)員擔(dān)任多個(gè)國內(nèi)外期刊審稿人，多次獲得IEEETrans期刊年度最佳審稿人4中國科學(xué)院電工研究所INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGESWhyWhytheobservabilityooperationorfaultsin●Sensors'coveragerate(likeaccuracyoftopologyinformationcannotbegua●ThedevelopmentofDERs(ES,DG,controllableload,EV..,)hasexacerbatedthefrequencyoTopologyidentification(TI)isconductedtodetecttopologychangesdeventsandforsomecontrolandoptimizationfinctionsinnormaloperation.如如中國科學(xué)院電工研究所Technicalsupportforthetopologyide/blog/blog/use-case-mapping-and-planning-of-electric-distribution-network/中國科學(xué)院電工研究所/product/HslxrohZELKV/China-Advanced-Metering-Infrastructure-AMI-and-·MarkovRandomExistingstudiesintopologyidentificationMathematicaloptimizatMathematicaloptimizat·Mixed-integerline以以DalaviE,GolshanMEH,HatziargyriouND.Areviewontopologyidentificationmethodsanddistrbutionnetworks[J].ElectricPowerSystemsRescarch,2024,234:1●Synchronousphasormeasurement(e.g.μ-PMU)isnotwidelyavailable.中國科學(xué)院電工研究所INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGESSnge:HistorielDataldantiAtwo-stagetopologyidentificationframeworkforPDNsisproposedtorectopologesinhistoricalbatchdataandpredictthereal-timetopologybasedontheavailablenodalmeasurements.HistoricalUnlabeledDaSampleP:[spi,Vrl……Spm,ySampleP:[Spi,Ypi,Tprl……SpForeachrecord[sy,y],itisthenodameasurementnodesinthePDNatapointvoltageamplitudesandphaseangleswhiincludestheactiveanreactivepowerinjecategoryTiyisdeterminedprocedureisperformed,whichisatogeneratethelabeleddata[Sj,yj.Tg].Theparametersinthehistoricaltopologarithmiclikelihoodp(x?,Tj;O)istheprobabilityofdistributionofthej-thtopologyforthei-threcordinthesampleExpectation-maximization(Expectation-maximization(TheEMiterationalternatesbetweentheexpeandmaximizationstep(M-sTheE-stepcalculatesQt+1)(with1)inthematrix)baseparametersθ(t).Theconditionalcanalsoberegardedastheposteriorprobabrecordbelongingtothej-thtopologyTheM-stepmaximizesL(X,θ;Q(t+1))toupdatetheSSchematicdiagramSchematicdiagramoftheEMalgorithmintopomaynotbeavailableinpractice,thenamodifi22中國科學(xué)院電工研究所ProbabilityProbabilityDensityp(x;|Tj)p([si,y;]|Tj)p€=27_?Lp=abcBygif(vfpfcos(θ"-0f)-0evfcos(0-0F))-Bybifvevysin(O-0f)q=2=1ZB=abcBybi(vevcospfcs(P-09)-B,9Droy中國科學(xué)院電工研究所y'=h,(si)yi-8y=h;(si-E?)≈h;(Si)-ε?hAccordingtotheaboveexpressions,yi-h,(s?)~N(0,Z),Es~N(0,cs2),εy~N(whereZj=(h;(S)os)2+o,2.Thenp([s,y,s/,y:|7)canbereplacedbyp(y?-h,(s:|T),andthelatteronlycontainsmeasurementvaluesanp(x;|T;)=p(E;|Tj)=p(y;-h;(largeamountofhistoricaldatacanbelabeled.●Thentheclassifiersforthelabeledhistoricaldatausingaseriesof·Theclassifierisntopologiesthatdidnotexistinthehistor·Aclassifiermayalsohaveerrorbyits·Thecredibilityanalysisshould·AllthepossibletoadvanceintheBayesianrecursion·Afterenoughiterationsarepeconvergetoonetopologycategory中國科學(xué)院電工研究所中國科學(xué)院電工研究所INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGES·Thenodalvoltageangles(exceptforthefirstbus)andbus'spowerinjectionsinthePDNaregeneratedusingtheMatpowert15161718·Measurementerrorsandaddedtotheabovege0tegories(allofthemdatasample,i.e.,T1,T2,T3,T4,correspondingtoTCTCαC-41.7] —— 一StandarddeviationofthemeasurementerrorsVariationrangeofthelineparameters SensitivitySensitivityanalysisofthetopo·1000recordsofeachtopologyc·TheClassificationLearnerRecognitiontoolboxesinMATLABareadohehistoricaldata,theirlogarithmicprobamuchsmallerthanotherrecordandthetopologyidentificationresultApplicationinaApplicationinaThrIC0—25中國科學(xué)院電工研究所INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGESTheoverallprocessoftheapp11中國科學(xué)院電工研究所·tominimizethetotalnumberconnectinglinesbetweensu·thenumberofnodes&thenumofunknownstates中國科學(xué)院電工研究所ThedistributionofthetestingdataintheTheresultsofsplit-EMprocessforthewholearea中國科學(xué)院電工研究所TABLEXITHECOMPARISONOFREAL-TIMERESULTSOF135-BUSSYSTEM30INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGES電動(dòng)電動(dòng)乳用乳用ACDC線路阻抗負(fù)荷aKP?=ViZkeNVk[Cicos(θi)+BixsinKP②Q=V?ZkeN,Vk[Gisin(Gik)-Biac②99中國科學(xué)院電工研究所T4[011110010034中國科學(xué)院電工研究所INSUTTUTEOFHLGTRICALENGINBERNGGHNSEACADENYOFSGIENGESAmodifiedEMalgorithm:Spli·Applicabletodifferenttypesoftopologies,differentsystemmodels.·ApplicablewhenthenodalvoltagephaseanglesarenAtwo-stagetopologyidentificationframew·Thenumberofthetopology·Applicableforlarge-Thecalculationburdenisreduc中國科學(xué)院電工研究所·Jointparametersestimationand中國科學(xué)院電工研究所●L.Ma,L.Wang*,Z,Liu.Topologyide