基于图像特征挖掘和深度森林的配网电缆终端故障诊断方法  

Fault Diagnosis Method of Distribution Network Cable Terminal Based On Image Feature Mining and Deep Forest

作　　者：黎珞 曹伟东 余洋 LI Luo;CAO Weidong;YU Yang(Shenzhen Power Supply Co.,Ltd.,Longhua Power Supply Bureau,Shenzhen 518110,China;College of Electrical Engineering,Southwest Jiaotong University,Chengdu 611756,China;Hubei transportation planning and Design Institute Co.,Ltd.,Wuhan 430051,China)

机构地区：[1]深圳供电有限公司龙华供电局,广东深圳518110 [2]西南交通大学电气工程学院,四川成都611756 [3]湖北省交通规划设计院股份有限公司,湖北武汉430051

出　　处：《电工技术》2021年第9期155-157,162,共4页Electric Engineering

摘　　要：为及时、准确诊断出配网电缆终端早期、潜伏性故障,提出了一种基于图像特征挖掘和深度森林的配网电缆终端诊断方法,采用深度自编码器对局部放电图像进行深度特征挖掘,结合深度森林网络对典型终端故障进行诊断。通过现场实测数据验证,该方法对三种典型故障的诊断率均保持在90%以上,且识别时间在10 s内,能够快速、准确、高效地对配网电缆终端故障进行诊断,识别率和识别时间均优于传统故障方法。To diagnose the early and latent fault of distribution network cable terminal timely and accurately,a method of distribution network cable terminal diagnosis based on image feature mining and deep forest is proposed.The depth self encoder is used to mine the depth feature of partial discharge image,and the deep forest network is used to diagnose the typical terminal fault.Through field test data verification,the diagnosis rate of the three typical faults is maintained above 90%,and the identification time is within 10 seconds.It can diagnose the cable terminal fault of distribution network quickly,accurately and efficiently.The recognition rate and recognition time are better than the traditional fault methods.

关 键 词：配网电缆 图像特征 深度森林 故障诊断 

分 类 号：TM41[电气工程—电器]