Abstract: Aiming at the problem that the noise of arrester leakage current affects the equipment monitoring work, an improved SVD-SVMD denoising method for AC filter arrester leakage current is proposed. First, the original leakage current data of the arrester undergoes singular value decomposition, during which the differential energy spectrum of singular values is introduced to determine the optimal threshold. The data are then reconstructed based on this optimal threshold to extract the primary harmonic components. Next, the minimum envelope entropy is utilized to optimize the penalty factor for successive variational modal decomposition, which is then applied to decompose the preprocessed data. After the decomposition, the dominant signal components are reconstructed. Finally, an improved wavelet threshold is employed to remove the residual micro-noise, resulting in a denoised leakage current signal of the metal oxide arrester. The effectiveness of the proposed method is validated through simulations and engineering experimental results.

Key words: arrester of AC-filter, leakage current, denoising, singular value energy difference spectrum, wavelet threshold