The oscillation wave partial discharge test has been used more and more in the detection of cable insulation defects, but the background noise interference and discharge pulse matching affect the accuracy of partial discharge location. A partial discharge localization method based on ensemble empirical mode decomposition and Hilbert marginal spectral energy is proposed to locate cable insulation defects accurately under oscillating wave voltage. Firstly, the partial discharge pulse signal is extracted by moving window threshold method. Then, the ensemble empirical mode decomposition algorithm is used to decompose the signal, and Hilbert transform is applied to each component to obtain its marginal spectral energy value. The proportion of different eigen modal function components in the original signal marginal spectral energy value is selected as the characteristic quantity, the signal similarity is calculated to match the incident pulse signal and the reflected pulse signal, and finally the partial discharge location is carried out by the time domain reflection method. The vibration wave withstand test is carried out on 35 kV cable, and the detected partial discharge signal is located and calculated. The results show that the pro-posed method can realize the location of cable insulation defect under the condition of strong Gaussian white noise, and the average location error can reach 1.15%.