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
李小龙 1桑多香 2曹洪亮3
作者信息
- 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;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
- 折叠
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.
Abstract
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Key words
metal oxide arrester/fault diagnosis/self-organizing feature map/leakage current引用本文复制引用
出版年
2025
NETL
NSTL
万方数据