Abstract: The operating condition of turbine generator bearing is complicated, the fault rate is high, the maintenance cost and the power generation loss are great. A signal noise reduction method based on Variational Mode Decomposition algorithm and compound fuzzy entropy is proposed to solve the problem of data screening of turbine bearing vibration signal. Firstly, the fuzzy entropy of each component after the original signal is decomposed by the VMD algorithm is calculated, and the components are eliminated successively according to the entropy value from large to small, and the fuzzy entropy of the remaining signal after the elimination is calculated to obtain the fuzzy entropy sequence. Three kinds of fault data of SL1500 wind turbine generator bearing are used to verify the ability of complex shannon entropy, compound fuzzy entropy and fault description. The experimental results show that the signal noise reduction method based on VMD algorithm and compound fuzzy entropy has good adaptability and decomposition effect, and can effectively eliminate the useless components in the signal, and has good noise reduction effect.