一种新金属氧化物避雷器故障在线诊断方法  

A New Method for Fault On-line Diagnosis of Metal Oxide Arrester

作　　者：李小龙 桑多香 曹洪亮 LI Xiao-long;SANG Duo-xiang;CAO Hong-liang(Hebei Meteorological Administrative and Technical Service Center,Shijiazhuang 050021,China;不详)

机构地区：[1]河北省气象行政技术服务中心,河北石家庄050021 [2]青海省海南藏族自治州气象局气象灾害防御中心,青海海南813000 [3]吴江市建设工程质量检验中心有限公司,江苏苏州215000 [4]南京信息工程大学,中国气象局气溶胶与云降水重点开放实验室,江苏南京210044

出　　处：《电力电子技术》2025年第1期25-29,共5页Power Electronics

基　　金：国家重点基础研究发展计划项目(973计划)(2014CB441405);河北省暴雨天气过程的闪电与降水关系研究(23ZC09)。

摘　　要：谐波失真和漏电流(尤其是阻性电流)是金属氧化物避雷器(MOA)故障诊断主要指标。针对传统MOA诊断方法测量参数多、操作难、正确率低等问题,基于格策尔和自组织神经网络(SOM)提出MOA故障诊断方法,通过提取MOA漏电流中基波及3,5,7,9次谐波幅值,训练SOM模型,实现MOA不同类型故障诊断。结果表明,训练后SOM模型对于MOA故障诊断正确率高达96%,计算耗时合理,兼顾了正确率和计算效率。Harmonic distortion and leakage current(especially resistive current)are the main fault diagnosis indexes of metal oxide arresters(MOA).Aiming at the problems of traditional MOA diagnosis method or the difficulty of multiple operation and low accuracy of measurement parameters,an MOA fault diagnosis method based on Goetzel and self-or-ganizing feature map(SOM)is proposed.By extracting the third,fifth,seventh and ninth harmonic amplitudes of the base wave in the MOA leakage current,SOM neural network model is trained to realize different types of MOA fault diagnosis.The results show that after training,the SOM neural network model has a correct diagnosis rate of up to 96%for MOA fault,and the calculation time is reasonable,both the correct rate and the calculation efficiency are taken into account.

关 键 词：金属氧化物避雷器 故障诊断 自组织神经网络 漏电流 

分 类 号：TM826[电气工程—高电压与绝缘技术]