XLPE cables, an important tool for power transmission in the carbon peaking and carbon neutrality goals, suffer from a decline in insulation performance after a number of years. Partial discharge (PD) detection has been widely used as an important means to evaluate the insulation status of XLPE cables. To solve the problem of various types of noise in PD signals, a PD signal denoising method based on Spearman variational mode decomposition (S_VMD) and spatial dependence recurrence sample entropy (Sdr_SampEn) is proposed. First, the signal is decomposed into K optimal intrinsic mode functions (IMF) through S_VMD. Secondly, the Sdr_SampEn value of each IMF is calculated to determine whether the IMF is the noised-dominated IMF or the PD-dominated IMF. After classification, the IMF will be denoised by improved wavelet threshold denoising and Savitzky-Golay (SG) filter denoising. Finally, the processed IMF will reconstructed, and the noise-reduced signal is acquired. This method is used to denoise the simulated and lab-measured PD signals, and it is further compared with many denoising algorithms such as adaptive variational mode decomposition (AVMD), etc. The results prove that the proposed method can effectively remove various types of noise, and thus shows engineering value. This work is supported by the National Natural Science Foundation of China (No. U1404522).