Intelligent Fault Diagnosis for Power Transformer Based on DGA Data Using Support Vector Machine (SVM)

Abstract

Transformer is a crucial element in distributing electricity from power plant. Disturbance in transformer operation should be avoided. Dissolved gas analysis (DGA) has been known as one of the most effective tools to monitor the health of transformer. There are various methods in interpreting DGA manually, such as IEEE and IEC-based methods. However, those methods still require the human expertise. Fast and accurate fault diagnosis in the transformer remains a challenge. This study proposes an intelligent system to diagnose fault types in the transformer using data mining approach, i.e. support vector machine (SVM). SVM has been known for its robustness, good generalization ability and unique global optimum solutions. IEC TC10 databases are used as data to illustrate the performance of multistage support vector machine (SVM). The proposed system yields effective transformer fault diagnosis with high recognition rate, which is around 90%.