Artificial neural networks and genetic algorithm for transformer winding/insulation faults

This paper presents an application of Artificial Neural Network and Genetic Algorithm for transformer winding/insulation faults diagnosed using Dissolved Gas in Oil Analysis. A back propagation training method is applied in neural network to detect the faults without cellulose involvement. Genetic Algorithm is used to derive the optimal key gas ratios to enhance the accuracy of fault detection. The dissolved gas in oil analysis method is known to be an early fault detection method and enables to carry out diagnosis during online operation of the transformer. Besides, the condition of the transformer could be monitored continuously by time to time. The results are compared between the real and predicted faults to observe the accuracy rate of the system.