基于超声波法的GIS绝缘缺陷类型识别  

A Ultrasonic Detection Based GIS Insulating Defect Types Recognition

作　　者：律方成[1] 张波[1] 

机构地区：[1]华北电力大学河北省输变电设备安全防御重点实验室,河北保定071003

出　　处：《电测与仪表》2014年第14期22-26,共5页Electrical Measurement & Instrumentation

基　　金：国家高技术发展计划项目(863计划)(2011AA05A121);中央高校基本科研业务费专项资金资助项目(13ZD14)

摘　　要：在气体绝缘组合电器(gas insulated switchgear,GIS)实体模型内部模拟了高压导体表面突起、悬浮金属颗粒和绝缘子表面固定金属颗粒三种绝缘缺陷,其中用针-板放电模型模拟高压导体表面突起缺陷。GIS模型内部充入0.4MPa的SF6气体,当加压到60kV时,三种模型均有稳定的放电。用超声波传感器分别测得其响应的放电波形100组,取相邻两个半波的信号幅值差的绝对值Udif和一个周波内的信号值的绝对值之和Utal作为特征量,用BP神经网络进行识别,识别率在80%左右,最后用最小距离分类器与BP神经网络的分类结果做对比,证明了BP神经网络的优越性。Three insulation detfects of high voltage conductor surface protrusion, suspended metal particles and insulated straping metal particles are simulated in gas insulated switchgear （GIS） entity model. A needle-plate discharge model has been adopted to simulate the high voltage conductor surface protrusion defect. 0.4MPa SF 6 gas has been inflated into GIS model. When the voltage is added to 60kV, these three models will all have stable discharge. Ultrasonic sensor is used to measure the discharge waveform for 100 groups. The absolute value Udif of the amplitude difference between the adjacent half waves, and the sum of the signal cycle absolute value Utal are takedn as the characteristic parameters. The defect types are recognized by BP neural network and the recognition rate is about 80%. Finally, the least distance classifier is used to compare with BP neural network. The results prove the superiority of BP neural network.

关 键 词：气体绝缘组合电器 超声波检测法 BP神经网络 绝缘缺陷类型 

分 类 号：TM855[电气工程—高电压与绝缘技术]