Application of Counter Propagation Network in Fault Diagnosis of Power Transformer

Abstract

Counter Propagation Network (CPN) is presented in this paper to diagnose the fault of power transformer. CPN have both respective advantage of Self-Organizing Map Network and Grossberg Self-Organizing Competitive Network, such as classification flexibility, simple algorithm and fine classification accuracy. In order to improve the classification accuracy, the conception of combination is introduced. The fault diagnosis of power transformer is consisted of 4 CPN. The first CPN is used to classify the normal and fault. The second CPN is used to classify the heat fault and partial discharge fault. The third CPN is used to classify overheating faults in magnetic circuit (MC) and overheating faults in electrical circuit (EC). The fourth CPN is used to classify discharge faults related to solid insulation (RSI) and discharge faults unrelated to solid insulation (USI). The example shows that the effect of combinatorial CPN is a good classifier in the fault diagnosis of power transformer.