Fault Identification in Power Transformers Using Dissolve Gas Analysis and Support Vector Machine

Abstract

Transformer faults need to be identified accurately at the early stage in order to ease the maintenance of power transformer, reduce the cost of maintenance, avoid severe damage on transformer and extend the lifespan of transformer. Dissolved Gas Analysis (DGA) is the most commonly used method to identify the transformer fault in power system. However, the existing transformer fault identification methods based on DGA have a limitation because each method is only suitable for certain conditions. Thus, in this work, one of the artificial intelligence techniques, which is Support Vector Machine (SVM), was applied to determine the power transformer fault type based on DGA data. The accuracy of the SVM was tested with different ratio of training and testing data. Comparison of the results from SVM with artificial neural network (ANN) was done to validate the performance of the system. It was found that fault identification in power transformers based on DGA data using SVM yields higher accuracy than ANN. Therefore, SVM can be recommended for the application of power transformer fault type identification in practice.