基于峭度和间谐波分析的故障电弧识别方法  

Arc Fault Identification Method Based on Kurtosis and Interharmonics Analysis

作　　者：崔芮华[1] 曹欢[1] 耿丽恺 刘斌 CUI Ruihua;CAO Huan;GENG Likai;LIU Bin(Electrical Research Institute of Hebei University of Technology,Hebei University of Technology,Tianjin 300130,China;Tianjin Research Institute of Electric Science Co.,Ltd.,Tianjin 300301,China)

机构地区：[1]河北工业大学河北工业大学电器研究所,天津300130 [2]天津电气科学研究院有限公司,天津300301

出　　处：《电气传动》2018年第10期79-84,共6页Electric Drive

基　　金：河北省自然科学基金青年基金(E2015202143);河北省教育厅青年基金(QN2014148)

摘　　要：为了对115 V/400 Hz供电条件下的线路进行串联电弧故障检测,计算了点接触试验中发生电弧故障前后电流信号的峭度值、间谐波数学期望和方差,以此作为识别电弧故障的指标和BP神经网络的输入,将线路故障和正常作为输出,采用带附加动量的学习方法对神经网络进行训练。该方法的电弧识别正确率在98%以上。In order to detect the series arc fault of the line which is under the condition of 115 V/400 Hz,the value of kurtosis,the mean and variance of interharmonics before and after the acr fault were calculated in the experiment of point contact.Regarded them as the indicators which identified the arc fault and the inputs of BP neural network,and the output was the normal or fault state of the line.The BP neural network was trained by the learning method with the additional momentum.The recognition of the test samples shows nearly 98 percent accuracy rate.

关 键 词：电弧故障 峭度 间谐波分析 误差反向传播神经网络 

分 类 号：TM501[电气工程—电器]