电工技术学报  2022, Vol. 37 Issue (9): 2388-2398    DOI: 10.19595/j.cnki.1000-6753.tces.210504
高电压与放电 |
基于点对称变换的乙丙橡胶电缆终端缺陷诊断
周利军1, 刘聪1, 权圣威1, 曹伟东1, 项恩新2
1.西南交通大学电气工程学院 成都 611756;
2.云南电网有限责任公司电力科学研究院 昆明 650217
Defect Diagnosis of EPR Cable Terminal Based on Symmetrized Dot Pattern
Zhou Lijun1, Liu Cong1, Quan Shengwei1, Cao Weidong1, Xiang Enxin2
1. College of Electrical Engineering Southwest Jiaotong University Chengdu 611756 China;
2. Electric Power Research Institute Yunnan Power Grid Co. Ltd Kunming 650217 China
全文: PDF (4033 KB)   HTML
输出: BibTeX | EndNote (RIS)      
摘要 为解决高速动车组车载电缆健康状况的智能化诊断受限于天窗期短的问题,该文提出一种基于点对称(SDP)的乙丙橡胶(EPR)电缆终端缺陷局部放电诊断方法。首先,搭建局部放电试验平台获取局部放电信号;然后,提出一种车载电缆局部放电信号的SDP参数确定方法,并基于SDP变换将不同类型缺陷局部放电信号映射到极坐标系中形成SDP图像;最后,对比三种常见的深度学习网络——卷积神经网络(CNN)、栈式自编码器(SAE)及深度置信网络(DBN)提取不同类型缺陷的SDP图像深层特征,并基于网络尾端Softmax分类器进行识别。结果表明:针对四种典型的电缆缺陷,DBN网络与SDP图像的结合效果最佳,缺陷识别率达到了96.1%,相比于传统诊断方法,识别准确率提高了10%左右,由此验证了通过深度学习算法自适应提取SDP图像特征的方法,可有效应用于电缆缺陷诊断领域,具有较好的工程应用前景。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
周利军
刘聪
权圣威
曹伟东
项恩新
关键词 乙丙橡胶深度学习对称点模式状态识别    
Abstract:In order to solve the problem that the intelligent diagnosis of the health condition of on-board cables of high-speed electric multiple units (EMU) in China is limited by the short skylight period, a partial discharge (PD) diagnosis method based on symmetrized dot pattern (SDP) for the terminal defects of ehglene propylene rubber (EPR) cables is proposed. Firstly, the partial discharge test platform was built to obtain the partial discharge signal. Then, a method to determine the SDP parameters of the on-board cable partial discharge signals was proposed, and based on SDP transformation, the partial discharge signals of different types of defects were mapped to the polar coordinate system to form SDP images. Finally, three common deep learning networks—convolutional neural network (CNN), stack autoencoder (SAE) and deep belief network (DBN) were compared to extract deep features of SDP images of different types of defects, and Softmax classifier at the end of the network was used to identify them. The results shows that: Aiming at four typical cable defect, DBN network combined with the SDP image effect is best, defect recognition rate reached 96.1%, compared with the traditional diagnosis methods, identification accuracy increase by about 10%, thus verified through deep learning algorithm of adaptive SDP image feature extracting method, can be effectively used in cable defect diagnosis, and it has a good prospect of engineering application.
Key wordsEthylene propylene rubber(EPR)    deep learning    symmetrized dot pattern(SDP)    state recognition   
收稿日期: 2021-04-13     
PACS: TM247  
基金资助:四川省科技计划(2020JDTD0009)和广东省基础与应用基础研究基金(2020B1515130001)资助项目
通讯作者: 周利军 男,1978年生,教授,博士生导师,研究方向为电气设备状态检测与故障诊断。E-mail:ljzhou10@163.com   
作者简介: 刘 聪 男,1998年生,硕士研究生,研究方向为电缆局部放电特性及故障诊断。E-mail:847008771@qq.com
引用本文:   
周利军, 刘聪, 权圣威, 曹伟东, 项恩新. 基于点对称变换的乙丙橡胶电缆终端缺陷诊断[J]. 电工技术学报, 2022, 37(9): 2388-2398. Zhou Lijun, Liu Cong, Quan Shengwei, Cao Weidong, Xiang Enxin. Defect Diagnosis of EPR Cable Terminal Based on Symmetrized Dot Pattern. Transactions of China Electrotechnical Society, 2022, 37(9): 2388-2398.
链接本文:  
https://dgjsxb.ces-transaction.com/CN/10.19595/j.cnki.1000-6753.tces.210504          https://dgjsxb.ces-transaction.com/CN/Y2022/V37/I9/2388