基于深度学习与特征信号感知的局部放电故障检测算法设计  

Design of partial discharge fault detection algorithm based on deep learning andfeature signal perception

作　　者：王亮 辛伟 南锐娟 WANG Liang;XIN Wei;NAN Ruijuan(State Grid Tongchuan Electric Power Company,Tongchuan 727031,China;Xi’an Weiguang Science and Technology Co.,Ltd.,Xi’an 710065,China)

机构地区：[1]国网铜川供电公司,陕西铜川727031 [2]西安卫光科技有限公司,陕西西安710065

出　　处：《电子设计工程》2024年第9期42-46,共5页Electronic Design Engineering

基　　金：国网铜川供电公司项目(5226TC220004)。

摘　　要：针对开关柜局部放电故障检测准确度较低的问题,提出了一种基于CNN-LSTM-Attention模型的局部放电故障检测算法。该算法利用开关柜暂态地电压与超声波等监测信号,分别提取时域和频域特征参数作为输入数据。同时,采用CNN网络实现对输入数据高维特征的提取,并使用LSTM网络提高了时序数据对局部放电检测结果的影响。通过引入Attention机制加强对重要数据的关注程度,从而提升局部放电的检测准确率。仿真算例结果表明,所提算法的平均检测准确率可达95.1%,且相比于CNN及CNN-LSTM算法分别提升了约17%和12%,证明了该算法可以提升配电网运维的智能化水平。Aiming at the low accuracy of partial discharge fault detection in switchgear,a partial discharge fault detection algorithm based on CNN-LSTM-Attention model is proposed.The algorithm extracts the time-domain and frequency-domain characteristic parameters as input data by using the monitoring signals such as the transient ground voltage and ultrasonic wave of the switch cabinet.The CNN network is used to extract the high-dimensional characteristics of the input data.The LSTM network is used to improve the impact of the time-series data on the partial discharge detection results.The attention to important data is strengthened by introducing the Attention mechanism,so as to improve the accuracy of the partial discharge detection as a whole.The simulation results show that the average detection accuracy of the proposed algorithm can reach 95.1%,which is 17%and 12%higher than that of CNN algorithm and CNN-LSTM algorithm,and greatly improves the intelligent level of distribution network operation and maintenance.

关 键 词：局部放电 开关柜 LSTM 注意力机制 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TN99[自动化与计算机技术—控制科学与工程]