A CNN-BiLSTM-Attention model of aging prediction for high-temperature superconducting cables

High-temperature superconducting (HTS) cables have emerged as a promising solution for short-distance, high-current power transmission and urban electrical networks. However, traditional aging evaluation methods are inadequate for the distinctive material characteristics of HTS conductors and their complex operating environments. In this study, we leverage operational data from an in-service HTS cable project to investigate the interdependencies among cable core temperature, nitrogen cooling system temperature, and flow rate under different operating conditions. The HTS cables’ degradation trajectory is further assessed and three deep learning algorithms are compared to forecast the remaining useful life of the HTS cable. The proposed convolutional neural networks with bidirectional long short-term memory and attention mechanisms (CNN-BiLSTM-Attention) model achieves a 40–60% reduction in prediction errors. This work offers insights into effective approaches for predicting the aging process of HTS cables.