Abstract: Aiming at the non-Gaussian and non-stationary characteristics of the wind turbine vibration signals,thispaper proposed a new fault diagnosis method based on the hybrid time-frequency analysis. This new method dealedwith the raw vibration signals by the parameter optimizied Morlet wavelet de-noising method,to filter the strongbackground noise interruption. Then the auto term window method was introduced to suppress the cross terms in thetime-frequency domain, to strengthen the useful fault features, to extract the fault features and to realize the diagno-sis. In the parameter optimiziation process,the cross validation method was introduced to optimize the Morlet wave-let parameters and the scale parameter. The auto term function is designed based on the smooth pseudo Wigner-Ville distribution (SPWVD) and two threshold processes were introduced to reduce the computation and enhancethe computing efficiency. The wind turbine vibration signal analysis proved that this new method can not only sup-press the noise interruption in the background,but also realize the fault diagnosis efficiently.