基于SGMD-LSTM的GIS局部放电故障诊断方法  

GIS partial discharge fault diagnosis method based on SGMD-LSTM

作　　者：张运 张超 张士勇 马鹏墀 杨光 丁浩 Zhang Yun;Zhang Chao;Zhang Shiyong;Ma Pengchi;Yang Guang;Ding Hao(Yancheng Power Supply Branch,State Grid Jiangsu Electric Power Co.,Ltd.,Yancheng 224000,China)

机构地区：[1]国网江苏省电力有限公司盐城供电分公司,江苏盐城224000

出　　处：《电子技术应用》2025年第2期58-63,共6页Application of Electronic Technique

基　　金：国网江苏省电力有限公司科技项目(J2023002)。

摘　　要：为准确对气体绝缘开关设备(GIS)局部放电进行故障诊断,提出一种基于辛几何模态分解(SGMD)与改进长短神经网络(LSTM)的故障诊断方法。引入SGMD对局部放电信号进行分解;对信号进行多维特征提取,构造时-频-熵值混合特征向量;通过鱼鹰-柯西变异的麻雀优化算法(Osprey-Cauchy-Sparrow Search Algorithm,OCSSA)对LSTM的隐含层节点数和学习率进行自适应寻优;最后使用OCSSA-LSTM进行局部放电识别。实验结果表明,OCSSA在收敛精度、速度上有较大提升,表现优异;与其他故障诊断模型对比,OCSSA-LSTM故障诊断模型准确率最高可达97.5%,对实际GIS运维数据也能准确识别。To accurately diagnose partial discharge faults in Gas Insulated Switchgear(GIS),a fault diagnosis method based on Symplectic Geometric Mode Decomposition(SGMD)and improved Long Short Term Memory(LSTM)is proposed.SGMD is introduced to decompose partial discharge signals.Multidimensional features are extracted from signals and a mixed time-frequency-entropy feature vector is constructed.The Osprey-Cauchy-Sparrow Search Algorithm(OCSSA)is used to adaptively optimize the number of hidden layer nodes and learning rate of LSTM.Finally,OCSSA-LSTM is used for partial discharge identification.The experimental results show that OCSSA has significant improvements in convergence accuracy and speed,and performs excellently.Compared with other fault diagnosis models,the accuracy of the OCSSA-LSTM fault diagnosis model can reach up to 97.5%,and it can also accurately identify actual GIS operation and maintenance data.

关 键 词：GIS SGMD OCSSA LSTM 局部放电 故障诊断 

分 类 号：TM855[电气工程—高电压与绝缘技术]