基于改进型主成分分析的电力变压器潜伏性故障诊断

Incipient fault diagnosis based on improved principal component analysis for power transformer

基于变压器油中溶解气体分析(DGA),提出采用改进型主成分分析(PCA)法对变压器内部潜伏性故障进行诊断。该方法不采用传统主成分分析的标准化方法,而是采用样本指标绝对值之和对样本指标值进行标准化处理,既消除各指标数值在数量级上的差异,又保持了各个样本间的信息差异特征;根据主成分的累计贡献率选取样本主成分,对样本主成分之间的欧氏距离进行聚类,判断变压器的故障类型。实例诊断表明,该方法能有效地提高变压器内部潜伏性故障诊断的准确率。

Based on the transformer DGA（Dissolved Gas Analysis）,an improved PCA（Principal Component Analysis） is proposed to diagnose the incipient fault of transformer. Different from the traditional PCA,it standardizes the sample indices with the sum of their absolute values,which eliminates the numeric magnitude difference between indices while keeps their information differences. The principal components are selected according to the cumulative contribution rate and the Euclidean distances between them are clustered to determine the fault state of transformer. Diagnosis instances show that,the proposed method effectively improves the diagnosis accuracy of transformer incipient faults.