The Phase Selection of Three-phase Arc Fault Based on Improved TCN-LSTM

Abstract

According to the increasingly severe situation of electricity safety. Aiming at the difficulty in extracting arc fault features in low-voltage three-phase systems and the inconspicuous phase distinction, a three-phase arc fault phase selection method based on improved LSTM-TCN neural network is proposed. Firstly, the structure and activation function of Long Short-Term Memory Network (LSTM) and Temporal Convolutional Network (TCN) are improved according to the experimental data. Finally, the feature vectors obtained by the improved LSTM and the improved TCN are fused to determine the arc fault and the difference. The experimental results show that the model structure proposed in this paper can effectively extract the characteristics of three-phase arc faults and phases. It is of great significance to industrial and household electrical safety.