A novel algorithm for power fault diagnosis based on wavelet entropy and D-S evidence theory

Fast and accurate fault diagnosis is the primary prerequisite for separating the faulty devices and restoring the power supply, therefore it is of great importance to develop an advanced diagnosis method to meet the power system requirements. With the perspective of information fusion, this paper proposes a novel algorithm for fault diagnosis in power system via the fusion of several different wavelet entropies. Wavelet entropy can extract the fault characteristic quickly and accurately because it combines together the advantages of Wavelet Transform and Shannon Entropy; however in some conditions it is not easy to reach a satisfying result with single wavelet because of the uncertainty and diversity of faults. Therefore, several different wavelet entropies are fused by the D-S evidence theory and the basic probability assignment is set up by a weighted average method based on norm, then a decision method based on the basic probability number is used to diagnose the faults. Simulations with EMTDC and MATLAB demonstrate that this diagnosis method can increase the supporting rate of faults and improve the accuracy and the real-time performance of fault diagnosis in power system. Results also show that the proposed algorithm is feasible and reliable for fault diagnosis.