摘要
准确识别气体绝缘组合电器(Gas insulated switchgear,GIS)内部绝缘缺陷,有助于评估局部放电(Partial discharge,PD)的危害,并为设备检修提供重要依据。然而,传统的单一监测方法存在信息利用不全、特征维度过高、识别率较低的缺点。为了克服这类问题,首先,在220 kV真型GIS上搭建了局部放电试验平台,设置了绝缘子污秽、绝缘子气隙、悬浮电极和金属突出物等4种典型缺陷模型,采用脉冲电流法和特高频传感器采集各类缺陷的局部放电信号,将PD信号基于时间分析模式和基于相位分析模式的融合决策引入提出的BP神经网络和改进DS证据理论相融合的诊断方法中,通过引入证据相融度,利用提出的诊断算法进行故障识别。结果表明所提识别方法能深入挖掘有效信息,具有一定的拒误诊能力,故障识别率高于96.7%,显著优于传统的BP神经网络。
Accurately identifying internal insulation defects in gas insulated switchgear(GIS)helps assess the hazards of partial discharge(PD)and provides essential guidance for equipment maintenance.However,the traditional single monitoring method has the disadvantages of incomplete information utilization,high feature dimensionality and low recognition rate.To overcome these problems,a partial discharge experimental platform is first built on a 220 kV GIS,and four typical defect models,such as insulator fouling,insulator air gap,floating electrode and metal protrusions,are set up.The partial discharge signals of various defects are collected by pulse current method and UHF method.The fusion decision of PD signal based on time resolved partial discharge and phase resolved partial discharge is introduced into the proposed diagnosis method of BP neural network and improved DS evidence theory.By introducing the degree of evidence fusion,the proposed diagnosis algorithm is used for fault identification.The results show that the proposed recognition method can deeply mine effective information and has ability to reject false diagnosis,so that the final overall recognition rate is higher than 96.7%,which is significantly high than the traditional BP neural network.
作者
于聪
汤凯波
李哲
刘志鹏
陈博
刘远超
方雅琪
YU Cong;TANG Kaibo;LI Zhe;LIU Zhipeng;CHEN Bo;LIU Yuanchao;FANG Yaqi(Extra High Voltage Company,State Grid Hubei Electric Power Co.,Ltd.,Wuhan 430050;Hubei Engineering Research Center for Safety Monitoring of New Energy and Power Grid Equipment,Hubei University of Technology,Wuhan 430068)
出处
《电气工程学报》
CSCD
2023年第4期361-369,共9页
Journal of Electrical Engineering
基金
国网湖北省电力有限公司科技资助项目(521520220003)。
