Fault Diagnosis of Transformer Based on Random Forest

Abstract

Fault diagnosis of transformer in power system is studied in this paper. Considering the excellent performances of Random Forest (RF) in pattern recognition, we apply RF to construct a diagnosis model to predict the situation of transformer. The experiments of fault diagnosis for some real transformers show that RF obtains a better result in prediction accuracy and stability than traditional Back Propagation neural network does. In addition, the order of influence factors given by RF is helpful in fault diagnosis.