Single-Ended Transmission Line Fault Location Based on Multi-Wavelet Time–Frequency Feature Fusion and a Residual Neural Network

Accurate fault location on transmission lines is essential for rapid fault isolation and reliable power-system operation. This paper presents a single-ended transmission line fault location method based on multi-wavelet time–frequency feature fusion and a residual neural network. First, a complex-domain transient analytical model accounting for the frequency-dependent characteristics of transmission lines is developed to efficiently generate fault-induced traveling-wave samples. The continuous wavelet transform (CWT) is then applied to these samples using multiple mother wavelets, and the resulting time–frequency maps are stacked as three input channels to combine temporal localization with frequency-domain information. A dual-task residual neural network with ResNet-50 as its backbone subsequently performs fault-section identification and fault-distance estimation. PSCAD/EMTDC simulation results show that the proposed method achieves high fault location accuracy and good noise robustness. Further Frobenius-distance sensitivity and integrated gradients (IGs) analyses demonstrate the complementarity of wavelet features with different time–frequency resolutions, confirming the effectiveness of the multi-wavelet fusion strategy.