Fault Diagnostic Method of Power Transformers Based on Fuzzy CMAC Neural Network

Abstract

Dissolved gas analysis (DGA) plays an important role in fault diagnosis of power transformers. A novel diagnosis method based on fuzzy CMAC neural network (FCMAC) is proposed in this paper. The proposed fuzzy CMAC neural network has an optimization mechanism to ensure high diagnosis accuracy. The basis functions in the original CMAC are replaced with membership functions of fuzzy theory for smoothing the networks output and increasing the approximation ability in function approximation. A structure of the FCMAC with membership functions of different receptive fields is employed. These receptive fields are determined by the distributions of training data. So, the proposed structure can reduce the memory requirement a great deal in the original CMAC, and keep the same performance with the original CMAC. This proposed neural network has been tested by lots of real fault samples, and its results are compared with those of IEC ratio codes and CMAC neural network, which indicates that the proposed approach has remarkable diagnosis accuracy, and with it multiple incipient faults can be classified effectively.