Transformer Fault Portfolio Diagnosis Based on the Combination of the Multiple Bayesian Classifier and SVM

Due to the information of test data is incomplete and deviated in the power transformer fault diagnosis, and the Bayesian network can deal with uncertainty well. The article discusses the NB (naive Bayesian classifier), SB (selective Bayesian classifier), TAN (tree augmented naive Bayesian), BAN (BN augmented naive Bayesian classifier) and GBN (general Bayesian network), the five Bayesian classifier models for transformer fault diagnosis, and it is proposed a new method that the combination of the multiple Bayesian network classifiers and SVM for transformer fault diagnosis. The experiments show the portfolio model that that combined of multiple Bayesian classifiers and SVM is more suitable for transformer fault diagnosis, with a capacity processing the lack of information and more fault-tolerant performance, its performance is superior to single classifier method of diagnosis.