The low signaltonoise ratio of leakage vibration signal in watersupply pipelines (WSP) may result in the large leak location error. To address this problem, a leak location method based on the improved empirical wavelet transform (EWT) and crosspower phase difference spectrum is proposed. Firstly, the local minimum of wavelet packet energy spectrum is used to segment the signal spectrum. The signal energy bands of different scales are achieved through wavelet packet decomposition. The distribution of energy value is used to determine spectrum segmentation interval, which solves the spectrum division problem of EWT. Then, the leakage vibroacoustic signal is decomposed into multiple components by EWT according to the spectrum segmentation interval. Effective components can be selected according to the correlation coefficients, and the effective components are bandpass filtered using the frequency band with horizontal variation of the crosspower phase difference spectrum of the effective components. Finally, the leakage can be determined by crosscorrelation timedelay estimation of the filtered effective component signal. Compared with crosscorrelation, variational mode decomposition and crossspectrum analysis, simulation and experimental results show that the proposed method can effectively locate the leak point, and the average relative location error is reduced by 67 times and 15 times respectively.