Research on Topology of Artificial Neural Network for Transformer Fault Diagnosis

This paper describes the application of ANN(artificial neural network) in transformer fault diagnosis technology.We firstly introduce an ANN with fixed transfer function,and a method of ascertaining the optimum number of hidden layer neu rons.The paper then implements an evolutionary algorithm of evolution topology of ANN with multiple layer,variable transfer function and variable connections based on genetic algorithms.This algorithm can automatically find optimization topology of ANN of transformer fault diagnosis.The experiment results show that the algorithm has a good performance.