In order to improve the power quality disturbance classification accuracy, aiming at the characteristics of timing, a power quality disturbance classification method based on convolution-long short-term memory is used. Firstly, the disturbance signal is sampled as input. Secondly, the feature data is extracted via convolutional neural network. Then the extracted feature data is used as input of long and short term memory network in the form of sequence, and the feature data is screened and updated. Finally, the output feature data is classified by learning. The simulation results show that the average classification accuracy of this method for power quality disturbance signals is 99.6%, which is better than the single CNN method and the single LSTM method. This work is supported by National Natural Science Foundation of China (No.61503240 and No.61603246).