In order to solve the problems of large actual power quality disturbance data and low accuracy in identifying multiple power quality disturbances, a method of power quality disturbance recognition based on an adaptive Kalman filter based on maximum likelihood (KF-ML) and a deep belief network is proposed. First, the adaptive Kalman filter based on maximum likelihood is used to denoise the original disturbance signal with noise. Then the deep belief network is used to train and recognize the clean signal so that the type of power quality disturbance can be recognized. Finally, it is tested on power quality disturbance signals under 20 types of different noise levels. From the simulation results, it can be seen that the method in this paper has a higher classification accuracy rate under different noise levels. This shows the effectiveness of the method and is very robust to noise. This work is supported by the National Natural Science Foundation of China (No. 51507015).