Partial Discharge Fault Detection Method of CNN-LSTM Based on Fusion Attention Mechanism

In order to solve the problem that the randomness and diversity of partial discharge phenomena lead to the low recognition rate of traditional fault identification methods, this paper proposes an integrating attention mechanism model based on the ability of Convolutional Neural Network (CNN) adaptive feature extraction and the ability of long short-term memory network (LSTM) to extract features in the time domain. The CNN network is built based on the residual structure, the hopping connection structure is added to improve the robustness of the network, the attention mechanism (SE-Block) is introduced in the feature fusion process to strengthen the learning of effective features, and the CNN and LSTM are connected in parallel to improve the model's ability to identify partial discharge. This paper conducted experiments using publicly available partial discharge datasets and compared the algorithm with LSTM and CNN algorithms to verify the effectiveness of the designed partial discharge fault detection model.