Abstract: In view of the problem that partial discharge pulse signal of high voltage cable in coal mine is easily interfered by underground noise and impacts on accuracy of insulation monitoring, wavelet entropy spatial correlation algorithm was proposed for partial discharge pulse signal denoising. In the algorithm, value of wavelet entropy is used as threshold of noise energy, and correlation between partial discharge pulse signal and noise is adjusted by changing threshold, and then the noise is distinguished and removed. The laboratory experiment and field test results show that the algorithm not only reduces the possibility of misjudging noise as a signal, but also can eliminate the interference in every scale coefficients well, keep the integrity of the partial discharge pulse signal well and improve the accuracy of on-line insulation monitoring of coal mine high voltage cable.