A Single-Ended Intelligent Fault Location Method for Flexible DC Overhead Line and Cable Hybrid Transmission Line

This paper proposes a single-ended intelligent fault location method to address the uneven wave impedance between flexible DC overhead lines and cables. Firstly, the wavelet packet decomposition is utilized to extract the frequency domain energy spectrum matrix of the fault traveling wave, which is combined with the time domain waveform to form a combined eigenvector as an input to the model. Then, the CNN-GRU neural network model with attention mechanism is proposed based on the framework of “firstly determining the fault zone, and then accurately locating the fault distance”. The time-frequency signal of the fault traveling wave is used to realize the precise location of the fault. Finally, the simulation results on PSCAD/EM$T$DC show that the model is well adapted to high resistance faults, noise interference, and other operating conditions.