染污玻璃绝缘子泄漏电流特性及其闪络电压预测  

Flashover Voltage Prediction of Polluted Glass Insulators Based on the Characteristics of Leakage Current

作　　者：赵世华[1] 蒋兴良[2] 张志劲[2] 胡建林[2] 

机构地区：[1]国网湖南省电力公司电力科学研究院,湖南省长沙市410007 [2]输配电装备及系统安全与新技术国家重点实验室(重庆大学),重庆市沙坪坝区400044

出　　处：《电网技术》2014年第2期440-447,共8页Power System Technology

基　　金：国家重点基础研究发展计划项目(973项目)(2009CB72-4501/502/503)~~

摘　　要：有效预测绝缘子闪络电压是防止发生污闪事故的重要手段,泄漏电流是分析检测绝缘子闪络电压的重要方法。在人工污秽实验室进行大量的试验,模拟运行电压下污秽度与相对湿度对泄漏电流的影响,从不同角度提取了能够反映绝缘子表面污秽度及相对湿度的4个泄漏电流特征量:泄漏电流脉冲幅值熵S、脉冲幅值Ih、能量比K及能量E,并得到它们之间的变化规律。提出基于广义回归神经网络(generalized regression neural network,GRNN)的绝缘子闪络电压预测模型,将4个特征量S、Ih、K、E及相对湿度作为GRNN模型的输入量,闪络电压作为GRNN模型的输出量。预测结果与试验结果对比分析可知,相对误差小于7.33%,表明提出的绝缘子闪络电压预测GRNN模型的预测结果与试验结果基本一致,能够有效地对绝缘子闪络电压进行预测。Effective prediction of the flashover voltage of insulators is an important approach to the prevention of pollution flashover accidents, and the leakage current （LC） is an important factor to analyze and detect the flashover voltage of insulators. In order to predict the flashover voltage of insulators and prevent pollution flashover accidents, firstly, a large number of artificial pollution tests were investigated under different contamination levels and different relative humidity （RH）. Secondly, based on the experimental data, four characteristics of the LC, namely the entropy of pulse amplitude （S）, the maximum pulse amplitude （Ih）, the energy ration （K） and the energy （E）, were extracted. They reflect jointly how severe the contamination level of insulators and the RH are from different perspectives. Thirdly, the variation laws between the four characteristics and the contamination level, RH, were obtained. Finally, based on the relationship among them and generalized regression neural network （GRNN）, the flashover voltage prediction GRNN model were presented, in which the four characteristics and the RH were used as the inputs of model, and the flashover voltage was used as the output of model. Comparison between prediction results and test results showed that relative errors are less than 7.33%. Therefore, the GRNN model is valid and reliable to predict the flashover voltage of insulators and it can provide a reliable guide for operators.

关 键 词：泄漏电流 污秽度 相对湿度 绝缘子 闪络电压预测 广义回归神经网络 

分 类 号：TM852[电气工程—高电压与绝缘技术]