A Methodology to Diagnose Transformer Faults Based on Principal Components Analysis and Artificial Neural Network

A methodology, combining the principal components analysis (PCA) and the artificial neural network (ANN), to diagnose transformer faults based on dissolved gas analysis (DGA) data is proposed. The main features extracted from DGA data by the PCA are used as the input of the ANN to train and test the classifier. The proposed methodology is tested with practical transformer fault data. The test results show that a higher diagnosis accuracy can be achieved by using the main features extracted by the PCA as compared to that by the original features or the optimal feature combinations obtained by feature selection methods.