Abstract: The chromatographic analysis of the power transformer oil dissolved gas is an important method for transformer fault diagnosis by which the operating state of the transformer and the potential transformer internal fault can be grasped indirectly.Artificial neural network has been applied in the power transformer fault diagnosis successfully,but the  large number of learning samples and the complicated input-output relationship will lead to a slow network convergence.To resolve the problem,this paper employ the wavelet neural network improved by using genetic algorithms  in power transformer fault diagnosis,thus overcoming the shortcomings of  local minima and slow convergence speed.

Key words: wavelet neural network, genetic algorithms, power transformer fault diagnosis