Single-Phase-to-Ground Fault Line Detection in Distribution System Based on CNN-GRU

In view of the difficulty of extracting the single-phase ground fault（SPGF） features and the low accuracy of existing detection methods a wavelet transform which is continuous(CWT) and Convolutional Neural Network-Gate Recurrent Unit (CNN-GRU) fault line detection method is proposed. First, the continuous wavelet transform is applied to the zero sequence transient current to get the corresponding time–frequency grayscale image. Then, the CNN adaptively gets the local features of the time–frequency grayscale image, and the GRU learns the context dependency from the local features learned in the CNN layer. Finally, The SoftMax layer completes the fault line detection The simulation results show that the line detection accuracy is 99.41% based on the method proposed above, which is more accurate compared with existing methods.