Aiming at the problem that the subsequent feature extraction and recognition of vibration signal of transformer on-load tap-changer are affected by the environmental noise, the denoising method based on the variational mode decomposition (VMD) and the wavelet packet threshold algorithm is proposed. First, the vibration signal is decomposed by VMD method into a series of modal components with narrow band and distinguishing center frequency. Then the modal components are processed by the wavelet packet threshold algorithm separately, and a composite evaluation index constituted by root-mean-square error, signal-to-noise ratio and smoothness is used to determine the optimal wavelet decomposition level in order to obtain the best denoising result. Finally the signal is restructured to get the denoised vibration signal. An experiment is conducted on the transformer on-load tap-changer simulation experimental platform and the denoising analysis on the collected vibration signal is processed, which shows the effect of the proposed method is better than that of several common denoising methods.