The invention relates to the field of electrical automation, and discloses a three-phase power quality disturbance detection method based on a convolutional neural network. The three-phase power quality disturbance detection method includes the following steps of (A) collecting three-phase power disturbance signal data, (B) performing data pre-processing on three-phase power disturbance signals toobtain three-phase power RGB pictures, (C) make training sets and test sets, (D) building a convolutional neural network model and (E) obtaining three-phase power quality disturbance detection results. The three-phase power quality disturbance detection method is high in efficiency, the influence of human subjective factors is reduced, the collected three-phase electrical energy signals are converted into the RGB pictures, signal characteristics are more compact, calculation amount is less, the characteristics of the disturbance signals are extracted intelligently through the convolutional neural network model to achieve the accurate classification of the disturbance signals, multiple disturbance signals can be detected at the same time, and the situation of a fault occurred between two-phase or three-phase can be detected.