Single-phase ground faults in distribution networks have a high probability of occurring, and it is difficult to find the fault point. Inefficient way to find the fault point through manual patrol. To address the above problems, the response model of single-phase ground faulted overhead distribution line based on high-voltage pulse injection is established, and an offline fault ranging method for overhead distribution lines based on high-voltage pulse impulse impact is proposed by multi-dimensional characterization of line current signals. The method uses a high-voltage pulse signal generator to inject DC pulse signals into the fault line, analyzes the line current waveforms in multiple dimensions, such as time domain and frequency domain, and then applies convolutional neural network (CNN) for training to achieve the distance measurement of the fault point, which solves the problem of time-consuming and laborious manual discrimination with low accuracy, and improves the fault offline localization system. Use MATLAB/Simulink for modeling simulation simultaneously. The simulation shows that the fault ranging scheme has high accuracy and stability.