A new method to diagnose faults for high voltage circuit breakers is presented in the paper based on the introduction of wavelet packet and characteristic entropy. Firstly

the vibration signal is decomposed by the threelayer wavelet packet

and eight signals of each junction at the third level are reconstructed. Secondly

the vector is extracted from the envelopes of reconstructed signals. Finally the Euclidean distance between the vectors of normal state and fault state is used as a characteristic parameter for fault diagnosis. Experiment results demonstrate that the proposed method can easily and accurately diagnose breaker faults and detect the state change of circuit breakers in time domain and frequency domain at the same time.