The fault patterns of the traction power supply system under new energy access are becoming increasingly complex and diverse. Furthermore, the characteristics of high impedance fault nonlinear faults are weak, and traditional protection methods often result in misjudgement or omission of judgement. In light of these challenges, this paper proposes a method of identifying high-impedance faults of the traction network based on continuous wavelet transform and convolutional neural network. By constructing signal samples of high impedance faults and other transient disturbances in the traction network, applying the continuous wavelet transform to transform the signals into two-dimensional time-frequency diagrams, and inputting them into a convolutional neural network, the identification of high-impedance faults and other transient disturbances is achieved. The experimental results demonstrate that the recognition accuracy of the method presented in this paper reaches 95%, exhibiting high recognition accuracy and anti-interference ability.