Partial Discharge (PD) signals serve as critical indicators for evaluating the insulation condition of high-voltage equipment. However, these signals are frequently corrupted by significant noise interference during field detection. To enhance the detection accuracy of PD signals, this paper proposes a novel method that integrates Improved Particle Swarm Optimization (IPSO), Variational Mode Decomposition (VMD), and wavelet threshold denoising to process noise in four-channel PD signals. The method employs IPSO to optimize VMD parameters (bandwidth constraint α and number of modes K), followed by wavelet threshold denoising applied to the decomposed Intrinsic Mode Function (IMF) components, culminating in effective signal reconstruction. Experimental results demonstrate that this approach substantially improves the Signal-to-Noise Ratio (SNR), reduces the Root Mean Square Error (RMSE), and achieves a correlation between the denoised and original signals ranging from 0.9770 to 0.9967, thereby validating its efficacy in PD signal denoising.