A new fault diagnosis method for power transformers

Dissolved gas analysis can be abbreviated as DGA, which is a powerful and good method we used earlier to detect the initial fault operation of power transformers. In order to improve the accuracy of fault diagnosis based on DGA, this paper proposes a new fault diagnosis method that combines fuzzy logic with Cerebellar Model Articulation Controller (CMAC) neural network. Fuzzy CMAC Neural Network (FCMAC NN) utilizes each method to extract fuzzy logic diagnostic rules from a large number of fault samples, and trains CMAC based on the extracted rules to optimize its network. The combination of fuzzy logic and CMAC has an optimization mechanism to ensure high diagnostic accuracy for all general fault types. The proposed FCMAC neural network has been tested on many actual fault samples and its results have been compared with those of IEC ratio codes and CMAC networks, indicating that the method has significant diagnostic accuracy.