A Data-Driven Framework Based on Deep Learning with Self-Attention Mechanism for Fault Detection in Power Distribution Networks

Fault detection in distribution networks is an important means of ensuring the safe operation of power grids. With the expansion of the scale of distribution networks, fault detection has become more challenging in the context of distribution network state monitoring. In order to accurately detect faults and issue timely warnings, a data-driven method based on deep learning with self-attention mechanism is proposed. The multiple convolution kernels of different sizes are utilized to extracted local features at different scales. At the same time, Bidirectional Long Short-Term Memory (BiLSTM) is applied to learn the positive and negative information in the current data in order to better capture the temporal characteristics of the data. To further enhance the feature-captured ability of the proposed method, a self-attention mechanism is used to adjust the importance of each feature. Finally, the effectiveness of the proposed method is verified by analyzing the measured current data of a substation distribution network feeder, the results show that the proposed method can accurately detect abnormal conditions in distribution network lines.