基于KPCA-SVM的变压器多源信息融合故障诊断研究OA北大核心

Transformer is one of the most important power transmission and transformation equipment in power sys-tem and its insulation status monitoring and fault identification are of great significance to the safe and stable opera-tion 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 construct-ing 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.