基于时空图卷积网络的有源配电网故障定位方法

打开文本图片集

中图分类号：TN92-34；TM773 文献标识码：A 文章编号：1004-373X（2025）20-0105-08

引用格式：，，，等.基于时空图卷积网络的有源配电网故障定位方法[J].现代电子技术，2025，48（20）：105-112.

Method of active distribution network fault location based on spatio-temporal GCN

XU Yanbin，GAO Xuejun， ZENG Xiangjun，WANG Can，YU Bo，LI Ruiling （CollgeofElectrical andNewEnergy，China ThreeGorgesUniversity，Yichang4430O2，China）

Abstract：Inallusiontothechallngeoffault localization inactive distributionnetworksduetocomplex topologyand variablepowerflowdirections，amethodofactivedistributionnetworkfaultlocalizationbasedonspatio-temporalgraph convolutionalnetwork（CN）isproposedbyombinghespatitemporalcoelationofdistributionnetworkfaultinforationThe modedecompositionisemployedtoobtainandreconstructvoltagecomponentswith richerfaultfeatures.The Hilberttransformis usedtoextractvoltageamplitude-frequencyfeatures，whichiscombinedwiththedistributionnetwork topologytoconstructgraph data.Thespatio-temporalconvolutional module withinthespatio-temporal GCNisused to extractandfuse amplitude-frequency featuresofthevoltage，andtheoutputlayercanprovidethefaultindexofeach nodeinthegraphtorealizethefaultlocation. Thesimulationresultsshow thattheproposedmethodcan demonstratehighreliabiltyand generalizationunderdiferent fault scenariosbylearningthespatio-temporalfeaturesoffaultinformationateachnodeofthedistributionnetwork，andcanexhibit strong robustness under data interference.

Keywords：activedistributionnetwork；spatio-temporalgraphconvolutional network；faultlocation;spatio-temporal correlation; mode decomposition; Hilbert transform

0 引言

有源配电网由于拓扑结构复杂、潮流方向多变、设备数量众多，故障率较高。（剩余11251字）