Fault Section Identification of Distribution Network based on 2D-CNN-LSTM and Information Fusion

Aiming at the problem of fault section identification when single-phase ground fault occurs in 10 kV distribution network, this paper proposes to use wavelet packet transform to preprocess zero-sequence voltage signal. Then, 2-dimensional convolutional neural network and long short-term memory network (2D CNN LSTM), are combined to extract local features and global features that help to distinguish fault section in turn, and softmax function is used to map features into discrimination results of 2D CNN LSTM for each section. Finally, a mean maximum evidence fusion method based on accuracy weight factor (w-MM) is proposed to complete final judgment. The simulation results show that the proposed network structure has anti-noise performance, fast convergence speed and strong robustness. In addition, accuracy of w-MM reaches 99.79 %, and it is robust.