The swing signal of the hydraulic turbine unit includes strong impulse noise and Gauss white noise due to the disturbance of the field environment and equipment. This results in the real effective signal being submerged which brings great difficulty for signal extraction. In the light of this problem，a swing signal denoising method that combines median filtering，wavelet threshold de-noising and Savitzky-Golay smoothing filter is proposed. First，the median filter is used to remove the impulse noise in the analysis signal. Then，the Savitzky-Golay smoothing filter and threshold denoising are used to remove the white noise components in approximate coefficient and detail coefficient after wavelet decomposition respectively. The simulation results show that the method can effectively increase the signal-to-noise ratio（SNR），local correlation index（LCI） and smoothness（S），and reduce the mean square error（MSE）. The experimental results show that the method can not only remove the noise in the swing signal of the water turbine，but also effectively retain the details of the real signal.