A Novel Neural Network Approach to Transformer Fault Diagnosis Based on Momentum-Embedded BP Neural Network Optimized by Genetic Algorithm and Fuzzy c-Means

Transformer is one of the critical equipments for electric power transmission and distribution, and its safety situation plays an important role on stability and security level of power systems. Therefore, the diagnosis of its abnormal situation has always drawn enormous attention from both domestic and international scholars. Dissolved gas analysis (DGA) is a widely used method in transformer fault diagnosis field. However, the conventional DGA is not well suitable for transformer fault diagnosis because transformer's structure is complex and operating environment is changeable. On the other hand, the back propagation (BP) neural network, frequently employed in related field, also has some inherent disadvantages, such as local optimization, over-fitting and difficulties in convergence. So simply integrating conventional DGA to BP is not a good approach for fault diagnosis. Moreover, disturbance or noises within the training data, which is unavoidable due to systematic errors, may greatly influence the accuracy of diagnosis model with the growing size of the data. Thus, in this study, we integrate a combination ratio of taking advantages of IEC and Doernenburg, instead of usual DGA, into genetic algorithm (GA) and fuzzy c-means clustering algorithm (FCM) optimized BP, successfully building a novel model which has not been reported previously. Our results show this model has a better diagnosis accuracy rate and generalization ability than other models, and FCM and GA can significantly overcome the disadvantages of training data and BP, offering the potential of implementation for real-time diagnosis systems.