Electric power transformer fault diagnosis using OLS based Radial Basis Function Neural Network

Dissolved gas analysis (DGA) is one of the most useful techniques to detect incipient faults in electric power transformers. However, precise identification of the fault is not an easy task due to the variability of gas data and operational factors. Radial Basis Function Neural Network (RBF NN) is used to diagnose power transformer fault in this paper. The training gas exemplar extracted from transformer oil is selected to form the RBF NN. The Orthogonal Least Squares (OLS) learning algorithm is used to optimize architecture and parameter of the RBF NN. The input of RBF-type neural network architecture. The fault diagnosis system using RBF NN with and without optimization of OLS are investigated. Simulation on intensive test exemplars shows the improved precision of the RBF NN optimized with OLS learning algorithm.