基于多维时频特征的新型配电系统单相接地故障定位方法  

Single-phase-to-ground Fault Section Location in New-type Distribution System Based on Multidimensional Time-frequency Distribution Characteristics

作　　者：鲁晓天 唐金锐[1,2] 尹昕 黄云辉 周克亮[1,2] 袁成清 LU Xiaotian;TANG Jinrui;YIN Xin;HUANG Yunhui;ZHOU Keliang;YUAN Chengqing(School of Automation,Wuhan University of Technology,Wuhan 430070,China;State Key Laboratory of Maritime Technology and Safety,Wuhan 430063,China;School of Transportation and Logistics Engineering,Wuhan University of Technology,Wuhan 430063,China)

机构地区：[1]武汉理工大学自动化学院,武汉430070 [2]水路交通控制全国重点实验室,武汉430063 [3]武汉理工大学交通与物流工程学院,武汉430063

出　　处：《高电压技术》2025年第2期903-914,I0038-I0042,共17页High Voltage Engineering

基　　金：国家重点研发计划(2021YFB2601602);国家自然科学基金(52337006)。

摘　　要：新型配电系统柔性消弧装置及定位技术均需充分挖掘相电流暂态特征来实现选相、选线和故障定位。针对此问题,对新型配电系统单相接地故障相电流暂态分布特性进行分析,提出了一种基于相电流多维时频分布特征差异的新型配电系统单相接地故障定位新方法。依据故障相电流故障暂态量与非故障相电流故障暂态量的差异性,通过灰色关联度算法完成故障选相;对各出线始端监测点以及疑似故障馈线分支监测点的相电流暂态波形进行26维多维时频特征的提取,通过经方差优化的t-分布近邻嵌入算法(variance-optimized t-distributed stochastic neighbor embedding,VTSNE)进行筛选和降维,并对处理后的特征数据进行基于密度的有噪空间聚类算法(density-based special clustering of application with noise,DBSCAN)聚类完成故障选线和故障区段定位。该方法在某绿色港口10 kV新型配电系统模型中得到验证,在不同故障初相角、不同过渡电阻等故障场景下均可准确可靠定位故障位置,对采样同步精度及采样频率要求低,易于工程实现。The fault characteristics of transient phase currents should be fully captured and analyzed for fault phase,fault feeder and fault section identification in new-type distribution system integrated with renewable energy sources.To solve this problem,the transient distribution characteristics of phase currents in new-type distribution system are analyzed when single-phase-to-ground fault occurs,and a novel method of single-phase-to-ground fault section location in new-type dis-tribution system is proposed and discussed.The proposed method is realized by using the difference of multidimensional time-frequency distribution characteristics of phase currents.According to the difference between the fault transient quan-tity of the fault phase current and that of the healthy phase current,the fault phase selection is completed by the grey correlation algorithm.The 26-dimensional multidimensional time-frequency features are extracted from the faulty phase current transient waveforms of the monitoring points at the beginning of each outlet and the branch monitoring points of the fault feeder.The VTSNE is used to screen and reduce the dimension,and the processed feature data are subjected to DBSCAN to complete fault line selection and fault section location.This method is verified in a 10 kV new-type distribu-tion system in some seaport,which satisfies the adaptability under different fault initial phase angles and different transition resistance faults,and is not limited by the sampling synchronization rate of monitoring points.

关 键 词：新型配电系统 故障定位 多维时频特征 t-SNE降维 DBSCAN聚类 

分 类 号：TM862[电气工程—高电压与绝缘技术]