Abstract: Aiming at the problem of less fault information and fast current rise and high diagnostic speed requirements under low-impedance faults, a fault diagnosis method based on relative wavelet energy with sliding window and support vector machine is proposed. Based on the acquisition of local current for wavelet decomposition, the proposed method calculates the relative sliding window energy of wavelet coefficients at different levels, so as to construct fault characteristic vectors with low dimensionality and high sensitivity.Combined with multi-classification support vector machine, it can realize rapid and accurate diagnosis of short circuit faults, ground faults and normal working conditions. Simulation tests results based on MATLAB/Simulink show that the proposed fault diagnosis scheme can quickly and accurately identify short circuit faults and ground faults of DC microgrids.

Key words: DC microgrid, fault diagnosis, wavelet transform, support vector machine