基于遗传BP神经网络的绝缘子泄漏电流预测  

Evaluating of Leakage Current of Insulators Based on the GA-BP Neural Network

作　　者：张友鹏[1] 伍亚萍[1] 赵珊鹏 

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《铁道学报》2016年第5期46-52,共7页Journal of the China Railway Society

基　　金：国家自然科学基金(51567014);铁道部科技研究开发计划(2012J007-C)

摘　　要：为了探索绝缘子泄漏电流与各因素之间的关系,提出用遗传算法优化BP神经网络建立绝缘子泄漏电流Ih预测模型。首先针对单片瓷绝缘子进行人工污秽试验,利用泄漏电流测量系统记录不同运行电压U、相对湿度RH及等值附盐密度ρESDD下泄漏电流波形并进行分析;其次以U、RH、ρESDD作为BP神经网络预测模型的输入变量,利用遗传算法的全局搜索能力获得BP神经网络初始权值和阈值,建立泄漏电流幅值预测模型,并通过部分试验数据进行验证。结果表明:相较于利用最小二乘法及BP神经网络预测泄漏电流幅值,遗传BP神经网络提高了预测的精度和准确性。In order to explore the relationship between leakage current of insulators and influence factors,the genetic algorithm to optimize the BP neural network （GA-BP）was proposed to establish the prediction model of insulator leakage current.Firstly,artificial pollution tests were carried out on single suspension insulators, while waveforms of leakage current were recorded and analyzed by a leakage current monitoring system under the conditions of different operating voltageU,different relative humidity RH and equivalent salt deposit densi-tyρESDD .Secondly,with U,RH andρESDD as the input variables of BP neural network predictive model,after the use of the global searching ability of genetic algorithm to obtain the initial weights and bias of the BP neural network,a leakage current amplitude prediction model was established and verified by certain experimental da-ta.The results showed that the GA-BP neural network can improve the precision and accuracy of the predic-tion,compared with the prediction of leakage current amplitude by the Least Square method and BP neural net-work method.

关 键 词：绝缘子 泄漏电流 相对湿度 等值附盐密度 作用电压 遗传BP神经网络算法 

分 类 号：U255.43[交通运输工程—道路与铁道工程]