基于Tamura-HOG纹理特征与矩特征融合的配网电缆终端故障诊断方法  

Fault Diagnosis Method for Distribution Network Cable Terminal Based on Fusion of Tamura-HOG Texture and Moment Features

作　　者：魏亚军 李开灿 董振 WEI Yajun;LI Kaican;DONG Zhen(Jining Power Supply Company,State Grid Shandong Electric Power Company,Jining 272000,China)

机构地区：[1]国网山东省电力公司济宁供电公司,济宁272000

出　　处：《电力系统及其自动化学报》2022年第9期153-158,共6页Proceedings of the CSU-EPSA

基　　金：国网山东省电力公司科技项目(5206061800AT)。

摘　　要：针对现有的配网电缆因局部放电随机且复杂导致故障诊断尚存不足的问题,本文提出了一种融合局部放电谱图的矩特征、Tamura纹理特征和方向梯度直方图特征的特征提取方法。该方法从形状特征、全局和局部方向纹理特征3方面更加全面地表达局部放电谱图特征,对局部放电谱图的矩特征、Tamura和方向梯度直方图纹理特征进行提取与融合,结合自适应提升算法、后馈神经网络和支持向量机算法均可以实现识别准确率较高的故障诊断效果。然后,将本文方法与后馈神经网络、支持向量机及卷积神经网络、栈式自编码器深度学习算法进行识别对比,结果表明,自适应提升算法的识别准确率最高、耗时最短,具有良好的通用性和稳定性,为配网电缆的故障诊断提供了新的方法。Due to the problem that the partial discharges of distribution network cables are random and complicated,the existing fault diagnosis method is still insufficient.Under this background,a feature extraction method is proposed in this paper,which combines the moment feature of phase resolved partial discharge(PRPD)spectrum,Tamura texture feature and histogram of oriented gradient(HOG)feature.Using this method,the PRPD spectrum feature is fully expressed from the aspects of the shape feature and the global and local direction texture features,and its moment feature and the Tamura and HOG texture features are extracted and fused.With the combination of the adaptive boosting(AdaBoost)algorithm,back propagation(BP)neural network and support vector machine(SVM)algorithm,the fault diagnosis effect with a high recognition accuracy can be realized.In addition,the recognition result of the proposed method is compared with those of the BP,SVM,convolutional neural networks(CNN)and stacked autoencoder(SAE)deep learning algorithms,and results show that the AdaBoost algorithm has the highest recognition accuracy,the shortest time-consumption,a good versatility and a good stability,providing a novel method for the fault diagnosis of distribution network cables.

关 键 词：配网电缆 矩特征 Tamura-方向梯度直方图纹理特征 自适应提升算法 故障诊断 

分 类 号：TM74[电气工程—电力系统及自动化]