Abstract: The diagnosis method for abnormal tripping of circuit breakers in distribution stations mainly extracts the time-domain characteristics of circuit breaker signals and combines threshold discrimination to achieve fault classification. Due to the difficulty in capturing transient energy changes in current signals when faults occur, the diagnostic accuracy is poor. Regarding this, a diagnosis of abnormal tripping faults in 10 kV distribution station circuit breakers based on VMD wavelet packet is proposed. Construct a variational problem using VMD method and perform multimodal decomposition on circuit breaker current signals. Quantify the energy distribution of each frequency band using wavelet packet decomposition, and calculate the energy entropy as the feature vector to transform the time-domain signal into a frequency-domain energy mutation representation. By using Gaussian kernel function to map energy entropy features to high-dimensional space, and selecting the maximum probability category as the output result, abnormal trip fault classification and diagnosis can be achieved. In the experiment, the diagnostic accuracy of the proposed method was tested. The final test results show that when using the proposed method for diagnosing abnormal faults in circuit breakers, the algorithm has a misjudgment rate of less than 5% and has a relatively ideal diagnostic effect.