Reflected traveling waves are generated by substation equipment, and the incident waves and reflected waves are overlapped during traveling wave signal measurement. To address these issues, a precise detection method of traveling waves based on S-transform and particle swarm optimization and generated regression neural network algorithm (PSO-GRNN) is proposed. Firstly, the S matrixes of the overlapped traveling wave signal and the real incident traveling wave signal are obtained by the S transform, respectively. Secondly, the S matrixes of the overlapped traveling wave signal and the real incident traveling wave signal are reconstructed in terms of dimensionality into a vector, which is used as the input and output of PSO-GRNN for training and learning, and the network model for separating the overlapped traveling wave signal is established. Finally, according to this model, the S matrix of the incident traveling wave signal is separated from the overlapped traveling wave signal, and the S-inverse transform is performed to obtain the pure incident traveling wave. The simulation results show that the separated incident traveling wave has higher steepness and more prominent time-frequency characteristics, which provides a new idea to improve the reliability of existing traveling wave protection and the accuracy of traveling wave positioning.