Research on Method of Mechanical State Characteristics Diagnosis Based on STFT and RVM for Transformer Winding

During grid operation, as the transformer winding operates frequently, the failure rate increases accordingly, and then the life of the transformer is threatened. In the past, regular maintenance and fault repair, which possess heavy workload, inefficiency and low measurement accuracy, are adopted for the transformer winding. With smart grid developing, the resulting drawbacks have become increasingly prominent. One mechanical state characteristics diagnosis method based on STFT and RVM for the transformer winding is proposed, monitoring transformer winding in the case of the transformer operating, and achieving automatic fault diagnosis. In the method, the vibration probes are arranged in three-dimensional space of the X, Y, and Z to realize monitoring of vibration signals generated from the operation of transformer winding. By de-noise processing, background noise in transformer winding is effectively removed. Short-time energy method and Short-time Fourier Transform based on Black-man window ensures the eigenvalue extraction of transformer winding accurately. One fault diagnosis model based on the RVM which employ FCM clustering and cross-validation method is built. The results show that model can effectively identify the normal state and the fault state, and the identification accuracy of each fault state is above 80%, which lays solid foundation for monitoring technology of transformer winding's mechanical condition in the power grid.