Abstract: To fully study the time and frequency characteristics of acoustic emission (AE) from hydraulic turbines in cavitation conditions, this paper conducts a series of cavitation tests on a model Francis turbine, collects AE signals and extracts their features in the form of wavelet packet energies in frequency bands, and analyzes their variation relations in different cavitation stages. The results show that with cavitation coefficient decreasing gradually, the operation condition of the turbine changes from no cavitation to incipient cavitation, then to fully developed cavitation stage. Accordingly, significant variations occur in each of the wavelet packet energies in the feature frequency bands of AE signals collected from the locations of the guide vane lever and the runner lower band. In this process, the energy increases quite slowly at initial stage, reaches a peak around incipient cavitation and decreases slightly, then increases violently at critical cavitation until a peak value much greater than the first one, and finally drops sharply.