主元分析在变压器故障检测与识别中的应用

APPLYING PRINCIPAL COMPONENT ANALYSIS IN TRANSFORMER FAULT DETECTION AND IDENTIFICATION

针对IEC三比值法进行变压器故障诊断存在缺编码、编码边界模糊和诊断率偏低的问题,提出了采用主元分析和重构贡献图的故障诊断方法。在建立的PCA(Principal Component Analysis)统计过程模型上,构建SPE、T2统计量和重构贡献指标对变压器故障进行检测,并分析了贡献图法和重构贡献法的故障诊断性能。仿真结果表明基于主元分析和重构贡献图的故障诊断方法对数据更敏感,能够有效弥补IEC三比值法的不足,提高故障诊断正确率。

The fault diagnosis approach using principal component analysis and reconstruction-based contribution plot is proposed in the paper to solve the problems of IEC three-ratio method in diagnosing transformer fault including the lack of coding,faint boundary code and low diagnosis rate.Transformer fault detection is conducted through constructing SPE and T2statistical indices and reconstructing contribution indices based on the built PCA statistical process model,and the diagnosis performances of the traditional contribution plot method and the reconstructed contribution method are analysed as well.Simulation results show that the proposed method is more sensitive to the data and is able to effectively make up the deficiencies of IEC three-ratio method as well as to improve the correctness rate of fault diagnosis.