基于KPCA-SVM的变压器多源信息融合故障诊断研究  

Research on Multi-source Information Fusion Fault Diagnosis of Transformer Based on KPCA-SVM

作　　者：杨旭 周文 程林 罗传仙[1,2] 张静 江翼[1,2] YANG Xu;ZHOU Wen;CHENG Lin;LUO Chuanxian;ZHNAG Jing;JIANG Yi(NARI Group Corporation(State Grid Electric Power Research Institute),Nanjing 211006,China;Wuhan NARI Co.,Ltd.,State Grid Electric Power Research Institute,Wuhan 430074,China;State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830000,China)

机构地区：[1]南瑞集团(国网电力科学研究院)有限公司,南京211006 [2]国网电力科学研究院武汉南瑞有限责任公司,武汉30074 [3]国网新疆电力有限公司检修公司,乌鲁木齐830000

出　　处：《高压电器》2025年第2期54-62,共9页High Voltage Apparatus

基　　金：国家电网有限公司总部管理科技项目(超、特高压变压器油纸绝缘快速发展型故障检测与诊断关键技术研究)。

摘　　要：变压器作为电力系统中最主要的输变电设备之一,其绝缘状态监测与故障识别对于电力系统安全稳定的运行具有重要意义。变压器内部绝缘劣化所产生的局部放电信号是目前对其内绝缘状态评判以及故障类型识别的最为有效的判据之一。文中通过构建4种典型的变压器绝缘缺陷模型,搭建试验平台测量得到了局部放电下的特高频信号和超声波信号。通过对2种信号的分析,对于特高频信号在TRTD模式下和PRPD模式下提取了两组特征参量,对于超声波信号在TRTD模式下提取了一组特征参量。经过核主成分分析联合支持向量机(KPCA-SVM)的特征融合方法进行信息融合后,发现相较于单一信息的识别效果,信息融合的识别率有显著的提升。Transformer is one of the most important power transmission and transformation equipment in power system and its insulation status monitoring and fault identification are of great significance to the safe and stable operation of power system.The partial discharge signal generated by the internal insulation deterioration of the transformer is currently one of the most effective criteria for evaluating the internal insulation status and identifying the fault type.In this paper,the ultrahigh frequency signal and ultrasonic signal under partial discharge are obtained by constructing four typical insulation defect models of transformer and setting up test platform for measurement.Throughout the analysis of two kinds of signals,two sets of characteristic parameters are extracted for ultrahigh frequency signals in both TRTD and PRPD modes,and a set of characteristic parameters are extracted for ultrasonic signals in TRTD mode.It is found after the feature fusion method of kernel principal component analysis together with support vector machine(KPCA-SVM)for information fusion that the recognition rate of information fusion is significantly improved compared to that of single information.

关 键 词：变压器 特高频 超声波 信息融合 故障识别 

分 类 号：TM855[电气工程—高电压与绝缘技术]