The invention discloses a maximum entropy-spectrum method applied to disconnecting switch vibration monitoring.Measured data except for limited data is not subjected to any assumption of definiteness, solving is carried out through an extrapolation method according to data except for known limited data self-correlation sequences on the premise of the maximum information entropy, and the power spectral density of signals to be detected is estimated; consistent data extrapolation is carried out on observed data through an AR prediction model, a modern spectrum estimation method based on maximum entropy-spectrum estimation is adopted for carrying out frequency spectrum estimation on non-stationary signals of a disconnecting switch, and the analysis precision and accuracy of the method are further improved.By means of maximum entropy-spectrum estimation analysis on the vibration signals of the actual disconnecting switch, small sample data is adopted under the condition of not changing the sampling frequency of the signals, consistent data extrapolation is carried out on observed data through the AR prediction model, and the frequency resolution rate higher than classic spectrum estimation and spectrum estimation accuracy are obtained and adopted as a theoretical basis for adjusting the disconnected and connected state of the disconnecting switch.