Transformer Fault Diagnosis Study Based on Bayesian Case Library

Abstract

The transformer fault diagnostic method based on Bayesian case library is proposed in the paper. Firstly, a transformer fault case library is established by collecting standard guideline and expert experience. Secondly, by standardizing the states and fault modes in the case library, the method takes the states as inputs and the fault modes as outputs, which are used to train a naive Bayesian network classifier. When it is necessary for a fault diagnosis, the user is expected to input the fault states in order to finalize the correct fault mode with the help of the well-trained classifier. On this basis, and with the details of the fault mode, the method could help to get the fault diagnosis results of the case. Finally, the feasibility and effectiveness of this developed method is illustrated by a numerical example of transformer fault diagnosis on site.