Research on Fault Diagnosis of Superconducting Cable Based on SVM-BiLSTM-ATT

The identification of abnormal states in superconducting cables is crucial for ensuring the safe operation of power grids. In this paper, monitoring data of superconducting cables under both normal and abnormal conditions were collected. Based on these data, a Support Vector Machine (SVM) method was employed to extract features from the current data. Subsequently, the constructed feature dataset was input into a Bidirectional Long Short-Term Memory network with Attention (BiLSTM-ATT) model for training. The trained BiLSTM-ATT model was then compared with Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BiLSTM) models through comparative experiments to evaluate the accuracy of each model in diagnosing abnormalities during the operation of superconducting cables. The experimental results demonstrate that, compared with traditional LSTM and BiLSTM models, the BiLSTM-ATT model achieves the optimal diagnostic performance, with a model accuracy rate of 98.77%.