Aiming at the problems of multiple power quality disturbance signals, slow recognition speed, and complicated recognition process, a power quality disturbance signal recognition and classification method based on compressed sensing theory and one-dimensional convolutional neural network is proposed. This method uses discrete Fourier transform and Gaussian matrix to obtain the sparse vector of the original disturbance signal, uses the orthogonal matching pursuit algorithm to reconstruct the disturbance signal, and inputs the original disturbance signal and the sparse vector into the one-dimensional convolutional neural network classification model. It can be seen from the simulation results that this method can fully reduce the data volume of the disturbance signal to be processed by the existing recognition method, and realize the expression of the characteristic information of the disturbance signal with a small amount of data. It has high recognition rate for 14 types of single and compound disturbance signals with and without noise, which shows that the method has the characteristics of less sampling data, convenient feature extraction, high recognition rate and better noise robustness.