基于油中溶解气体分析的ISSA优化LSSVM变压器故障诊断研究  

Research on transformer fault diagnosis based on analysis of dissolved gas in oil of ISSA-LSSVM

作　　者：李雷军 吴超 付华[1] 齐致 王久阳 LI Leijun;WU Chao;FU Hua;QI Zhi;WANG Jiuyang(Faculty of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105,China;Institute of Electrical Engineering,Chinese Academy of Sciences,Beijing 100190,China;State Grid Huludao Power Supply Company,Huludao 125105,China)

机构地区：[1]辽宁工程技术大学电气与控制工程学院,辽宁葫芦岛125105 [2]中国科学院电工研究所,北京100190 [3]国网葫芦岛供电公司,辽宁葫芦岛125105

出　　处：《电工电能新技术》2023年第10期84-94,共11页Advanced Technology of Electrical Engineering and Energy

基　　金：国家自然科学基金项目(51974151、71771111);辽宁省高等学校国(境)外培养项目(2019GJWZD002);辽宁省教育厅科技项目(LJ2019QL015);辽宁省高等学校基本科研项目(LJKZ0352)。

摘　　要：为了提升变压器故障诊断的准确率,提出一种基于等规度映射(Isomap)与改进樽海鞘群算法(ISSA)优化最小二乘支持向量机(LSSVM)的变压器故障诊断方法。首先,基于油中溶解气体分析技术,构建14种能够反映变压器运行状态的故障特征,并结合Isomap对样本数据进行维数约减,消除变量信息之间的冗余数据;然后结合半数均匀初始化、混合反向学习策略和非线性递减权重因子策略对樽海鞘群算法(SSA)进行改进,并通过5个基准测试函数与原始SSA、粒子群算法(PSO)、正弦余弦算法(SCA)进行对比,证明其寻优能力和分类精度均有较大提高;最后用ISSA算法动态寻优LSSVM的惩罚参数γ和核函数参数σ,获取基于Isomap与ISSA-LSSVM相结合的故障诊断模型,并与PSO-LSSVM、SSA-LSSVM、SCA-LSSVM做对比实验,诊断精度分别为90.83%、81.67%、83.33%、80%。结果证明,所提方法能够有效地增强变压器故障诊断的性能。In order to improve the accuracy of transformer fault diagnosis,a transformer fault diagnosis method based on isometric mapping(Isomap)and improved salps swarm algorithm(ISSA)to optimize least square support vector machine(LSSVM)is proposed.First,based on the dissolved gas analysis technology in oil,14 types of fault characteristics that can reflect the operating state of the transformer are constructed,and combined with Isomap to reduce the dimensionality of the sample data and eliminate redundant data between variable information.Then the salp swarm algorithm(SSA)is improved by combining half uniform initialization,mixed reverse learning strategy and non-linear decreasing weight factor strategy,and compared with the original SSA,PSO and SCA algorithms through 5 benchmark test functions to prove its search.The optimization ability and classification accuracy have been greatly improved.Finally,the ISSA algorithm is used to dynamically optimize the penalty parameters and kernel function parameters of LSSVM,and obtain a fault diagnosis model based on the combination of Isomap and ISSA-LSSVM,and combine it with PSO-LSSVM,SSA-LSSVM,SCA-LSSVM for comparative experiments,the diagnostic accuracy is 90.83%,81.67%,83.33%,80%.The results prove that the proposed method can effectively enhance the performance of transformer fault diagnosis.

关 键 词：故障诊断 变压器 等规度映射 SSA算法 LSSVM 

分 类 号：TM41[电气工程—电器]