An improved Bi-LSTM identification method for cable partial discharge based on shrinkage residuals

Cable partial discharge is usually caused by cable defects. Effectively identifying local discharge signals can quickly determine the type of defects. A fault identification method for cable partial discharge based on improved shrinkage residuals and bidirectional long short term memory network is proposed. Firstly, an improved shrinkage residual is designed to filter the interference in the power sequence, and the spatial feature map without disturbance is extracted. Then, Bi-LSTM is used to extract sequence features, and bottleneck attention mechanism is introduced to enhance the attention to important features and improve the learning ability of the model, so as to achieve accurate fault identification. Finally, the experimental results show that the proposed method can achieve 98.87% recognition accuracy, which has stronger anti-interference ability and recognition accuracy than the traditional method.