Aiming at the problems of large amount of white noise and periodic narrow-band interference in partial discharge (PD) signal of mining high voltage XLPE cable, A denoising method based on improved Variational Modal Decomposition (VMD) is proposed for partial discharge signals of mining high-voltage cable. Firstly, the K in VMD algorithm is optimized by Spearman correlation coefficient, and then the noisy signal is decomposed into a series of intrinsic mode functions (IMF); IMF is divided into noised-dominated IMF and the PD-dominated IMF by kurtosis criterion; The noised-dominated IMF is denoised by improved wavelet threshold, and the rough penalty method is used to smooth the PD-dominated IMF; The processed IMF is reconstructed and the denoised signal is finally obtained. The simulation and measured signal results show that this method can effectively remove the noise in PD signal. Compared with the complete ensemble empirical mode decomposition (CEEMD) algorithm, the signal-to-noise ratio is improved by 46.29% and the mean square error is reduced by 32.61%.