In order to guarantee the operation stability of high-voltage circuit-breaker, this paper extracts eigenvector of vibration signals based on mining of characteristic entropy of the wavelet packet obtained under the state of circuit breaker. Besides, the eigenvector of various fault samples is adopted as the input information of the BP network. After that, the author conducts training of the well-built BP network to lay the foundation for the building of the high-voltage circuit-breaker fault diagnosis model based on the BP neural network and the wavelet characteristic entropy extraction. At last, through diagnosis simulation analysis of various faults, results suggest that: after wavelet packet characteristic entropy extraction of vibration signals of various faults, the eigenvector thus obtained has a high discrimination; the comprehensive diagnosis accuracy of the model can reach 92.5% after diagnosis of different fault types through the trained BP network. Therefore, the high-voltage circuit-breaker fault diagnosis model put forward in this paper has a high fault discrimination accuracy and can make up for defects of periodic maintenance and inefficient maintenance of the circuit-breaker.