Latent faults diagnosis for shunt reactor based on PCCA

A new shunt reactor fault diagnosis method based on principal component cluster analysis (PCCA) is proposed. This method is to combine the principal component method with cluster analysis method to identify shunt reactor latent faults. The sample data adopts typical gas concentrations of fault types from new IEC TC 10 databases. The sample data and diagnostic data form the original matrix. Ten the number of the components will be selected. The principal components of the original matrix are constructed to reduce overlapping and redundant information. Calculating the distance between principal components of typical sample data and the diagnostic data realizes cluster groups classification. The smallest distance of principal components between the diagnostic data and the corresponding typical sample data indicates they belong to the same fault type. For the sake of comparison, Duval triangle method and IEC60599 three-ratio method are also adopted to diagnose the same data. Results of diagnosis instances show the effectiveness of the proposed method.