Hybrid version of MLP neural network for transformer fault diagnosis system

Abstract

This paper proposed an intelligent classification system to diagnose fault in oil insulator power transformer based on dissolved gas analysis (DGA). The system constructs using the application of hybrid version of standard multilayer perceptron (MLP) neural network called Hybrid Multilayer Perceptron (HMLP) neural network. The network is trained using modified recursive prediction error (MRPE) training algorithm. Performance analysis of the HMLP network is compared with standard MLP network trained using three different algorithms, i.e. Bayesian Regulation, Lavenberg-Marquardt and Gradient descent. The experiment result indicated that the HMLP network attains the best performance in the transformer fault diagnosis.