Transformer fault diagnosis based on factor analysis and gene expression programming

A kind of gene expression programming algorithm (GEP) based on factor analysis (FA) is proposed and used for transformer fault diagnosis in this paper. The application of factor analysis can reduce the dimension and the correlation of the feature vectors, so as to decrease the computational complexity of the diagnosis classifier and increase the training and test accuracy. 170 groups of the transformer DGA data which can reflect the variety of the faults without redundant are used to study and get the GEP classifiers, while the other 130 instances is diagnosed by the GEP classifiers. The result of the diagnosis is rather exacting, which has obviously higher diagnostic accuracy than the result obtained by using Bayesian classification and BP network.