Study On Transformer Internal Fault Diagnosis Based On EBFNN

Power transformer is one of the most important equipment of the power system.Its failure will seriously affect the security, stability and economic of the power system,therefore, it will have great practical significance to strengthen the study of the internal fault diagnosis method of power transformer,and then propose a more effective detection and diagnosis to avoid failure or latent defect.In this paper, the method for transformer fault diagnosis based on ellipsoidal basis function neural network (EBFNN) has been explored and studied.After learning the current situation and development direction of transformer fault diagnosis method home and abroad,the EBFNN has been focused studied in this paper. Based on sample data, structure of the network has been designed,and the method for the transformer fault diagnosis based on the structure has been proposed. The actual numerical example shows that this kind of method for fault diagnosis owns high rate of diagnostic accuracy,and not subject to noise.Then RBF neural network-based transformer fault diagnosis method and fuzzy neural network-based transformer fault diagnosis method were compared with the EBFNN-based transformer fault diagnosis method, simulation results show that three kinds of transformer fault diagnosis methods all have high accuracy of fault diagnosis, but for RBFNN, since ellipsoid basis function in EBFNN makes the input space divided more clearly, thus EBFNN-based transformer fault diagnosis method owns a higher rate of correct diagnosis;for fuzzy neural network,the EBFNN has higher rate of correct diagnosis,and the convergence speed is more fast.Followed by testing capacity of a single EBFNN to diagnose the double fault of transformer, in view of the unsatisfactory nature of diagnostic results, one modular EBFNN and two modular EBFNN were designed, numerical example shows that:one modular EBFNN and single EBFNN nearly have the same fault diagnosis results,it is not satisfactory;but the two modular EBF neural network can accurately diagnose a double fault and does not affect the high accuracy rate of single fault diagnosis, compared with the previous two kinds of forms,it owns better diagnostic performance.