When measuring partial discharge signals of high-voltage cables, noise components dominated by periodic narrowband interference and random white noise are mixed. In order to suppress the influence of noise components on the accuracy of partial discharge signal measurement, a partial discharge signal noise reduction method for high-voltage cable based on Empirical Wavelet Transform (EWT) is proposed. The noisy signal is decomposed using adaptive EWT. By calculating the kurtosis value of the modal component, the pulse signal can be located, and the effective feature components screened are reconstructed. Finally, the redundant noise of a reconstructed signal is removed using an improved threshold function to effectively suppress multiple noise. By conducting simulation comparison and field test, it is shown that the method in this paper can effectively suppress narrow-band interference and white noise, retain the effective information in partial discharge of high-voltage cable to a large extent, and perform stable noise reduction in different noise environments. This work is supported by Natural Science Foundation of Henan Province (No. 182300410280) and Science and Technology Research Project of Henan Province (No. 182102310936).