In order to identify the operation state of the transformer on-load tap-changer

this study first summarizes its operation state and fault characteristics

and uses the empirical mode（EEMD） to decompose the vibration signals generated during the operation of the tap-changer into Multiple intrinsic mode function components（IMF） are combined with Hilbert transform and energy entropy extraction to obtain feature vectors based on time-frequency analysis.The feature vectors are input into the BP neural network model optimized by the adaptive genetic algorithm（AGA） for fault identification

and the data simulation is performed.The feature vectors are compared with the feature vectors（PPDC） extracted after phase space reconstruction

and the results are verified under different network models.The recognition accuracy and convergence speed of the method are improved.The results show that when PPDC fault samples are used as the model input

the recognition accuracy of the BP neural network model before and after AGA algorithm optimization is 81.68% and 88.32%

and the number of convergence is 981 and 363.As the model input

the recognition accuracy of the BP neural network model before and after the AGA algorithm optimization is 91.66% and 96.68%

respectively

and the number of convergence is 349 and 159.The AGA algorithm can significantly improve the performance of the BP neural network model.It can be seen that the time-frequency feature extraction method can be combined with the AGA-BP neural network to realize the effective identification of the operating status of the on-load tap-changer.