Transformer fault diagnosis using dissolved gas analysis technology and Bayesian networks

Abstract

Bayesian model is developed for transformer faults diagnosis using dissolved gas in oil analysis. DGA (Dissolved Gas Analysis) is the traditional and conventional transformer fault diagnosis method, which mainly depends on the experience of operators and of the percentages of dissolved gases. In addition, the only measurement of the gases percentage is not sufficient to evaluate the equipment health. There are several cases where the proportions of dissolved gases remain trapped in the transformer. Regarding this uncertainty and in order to make decisions in a certain environment, the model developed in this study represents a powerful tool for decision making. In addition, one traditional method of DGA does not enable the diagnosis of all faults, for example the Rogers Ratio Method diagnose five faults only, but using the proposed Bayesian network (BN) it is possible to diagnose all faults from the same model. To illustrate the advantages of Bayesian methods in transformer fault diagnosis, a study of power station main transformer is conducted and the results are analyzed and discussed.