Mechanical vibration signals generated during high-voltage（HV） circuit breakers（CBs） operation contain abundant CB state information. However

due to the influence of noise

effective state information is often difficult to extract. To address this problem

a novel noise reduction method based on spectral shape variational mode decomposition（SSVMD） and fast singular value decomposition（FSVD） is proposed. Firstly

the optimal number of decomposition layer of the vibration signal is determined according to the spectrum characteristics of the variational mode decomposition（VMD） results

which realizes the adaptive decomposition of the vibration signal. Then

the power spectrum of the vibration signal is calculated to determine the noise frequency band. Finally

FSVD and signal reconstruction are carried out according to the main frequency components of the noise band for realizing the adaptive noise reduction. The algorithm proposed in this paper avoids the blindness of the choice of VMD layer number and singular value decomposition（SVD）effective reconstruction rank

and can perform adaptive noise reduction according to the characteristics of vibration signal.The effectiveness and practicability of the proposed noise reduction method are verified from qualitative and quantitative aspects through simulation signal and vibration signals of two different types of CBs. The results show that the proposed method has better noise reduction performance than the denoising methods such as median filtering

five-point cubic smoothing

wavelet denoising and empirical mode decomposition（EMD）

which provides a more suitable approach for noise reduction of CB vibration signal.