The invention discloses a single-phase earth fault line selection method for a power distribution network based on time-frequency image information. The method is characterized in that zero-sequence current waveforms of the head ends of lines after occurrence of a single-phase earth fault are generated through an established power distribution network model, and corresponding time-frequency imagesare obtained through corresponding conversions, so that a corresponding training set is obtained; and a convolutional neural network is trained through the training set to obtain a fault line selection model, fault time-frequency images are generated again through the same power distribution network model to form a testing set, the images are input into the convolutional neural network to obtainnetwork output values corresponding to the lines, and the line with the maximum output value is selected as a fault line, so that good line selection effects are finally achieved. The method disclosedby the invention has the advantages that the fault line can be accurately selected when the single-phase earth fault in the power distribution network occurs, so that great significance is achieved for fault diagnosis and reliable operation of the power distribution network with a complicated and changeable structure and weak fault signals.