Comparative Research Of Artificial Neural Network Applied In Transformer Fault Diagnosis

The transformer fault diagnosis is very important to the safety and economic operation of power system. Dissolved Gas Analysis (DGA) is one of the most effective methods to predict and diagnose transformer faults.In this thesis, on the foundation of collecting the transformer fault data widely, Three-Ratio Method is analyzed detailedly. Three-Ratio Method has high exactness rate, but it may bring about mistakes if the ratio is just at the boundary of the coding-zone.The neural network has the information processing traits of associative memory and strong capability to recognize and classify the input samples. The possibility of the practical application of artificial neural network to diagnose fault of equipments is come true.In this thesis, three kinds different of artificial neural network were applied in the transformer fault diagnosis, such as the BP network, the GA-BP network and the RBF network. The structure and principle of different network is introduced separately. In the different parameter selecting simulation process, the influence of the network training effect and extensive ability is expounded. The fault diagnosis is realized by MATLAB programming. The massive empirical data result analysis indicated that, the RBF neural network is exuding the ability and the training speed obviously compared to other two kinds of network.