Application of a combinatorial neural network model based on cluster analysis in transformer fault diagnosis

Abstract

The multi-resolution identification of transformer faults is significant for the maintenance of transformer. In this paper, a combinatorial artificial neural network (ANN) model, based on cluster analysis of data of dissolved gases in transformer oil, is presented. A more detailed classification is necessary to obtain explicit diagnosis results. Based on the discussion of traditional classification methods, a twelve-fault classification method is established. However there are similarities among these faults. which should be considered before constructing the combinatorial model. Hence, hierachical cluster analysis is chosen to investigate the similarities and helps to construct the model. Finally, the application results show the value of this model for the diagnosis of transformer faults.