基于神经网络深度学习的电缆故障异常分析与特征分类算法  

Cable Fault Anomaly Analysis and Feature Classification Algorithm Based on Neural Network Deep Learning

作　　者：黄烈江 沈狄龙 夏明明[1] 付凯 吕渭 HUANG Liejiang;SHEN Dilong;XIA Mingming;FU Kai;LV Wei

机构地区：[1]杭州欣美成套电器制造有限公司,浙江杭州310000 [2]杭州宇嘉微科技有限公司,浙江杭州310000

出　　处：《电力系统装备》2024年第9期40-42,共3页Electric Power System Equipment

摘　　要：文章是以解决电缆故障异常分析为目的,以卷积神经网络模型和特征分类算法为基础进行的研究。分析了电缆故障的类型和原因,并收集了来自多个电力系统的电缆运行数据说明处理步骤。分析深度学习模型和特征分类算法的构建和训练,解释技术细节和参数调优,以达到最好的分类性能。同时,设计并利用特征分类算法从时域到频域,采用支持向量机(SVM)和随机森林对数据进行分类。实验结果表明所提出的方法在电缆故障异常分析中表现良好,有高度的自动化优越性能和强泛化能力,为电缆故障异常分析提供了一种高效和精确的方法。In order to solve the cable fault anomaly analysis,this paper studies the convolutional neural network model and feature classification algorithm.The types and causes of cable faults are analyzed in detail,and cable operation data from multiple power systems are collected for processing.In the construction and training of deep learning model and feature classification algorithm,the technical details and parameters are optimized to achieve the best classification performance.At the same time,the feature classification algorithm is designed and used to classify the data from time domain to frequency domain by SVM random forest.The experimental results show that the proposed method performs well in the analysis of cable fault anomalies,and has a high degree of superior automation performance and strong generalization ability.Through the research of this paper,an efficient and accurate method is provided for the analysis of cable fault anomalies.

关 键 词：电缆故障 深度学习 卷积神经网络 特征分类 

分 类 号：TM855[电气工程—高电压与绝缘技术]