Quench Prediction for REBCO Pancake Coils Using LSTM

In this article, we propose a neural network-based quench prediction method. High-temperature superconductors (HTS) has a slower propagation velocity in the local normal-zone than low-temperature superconductors (LTS), and the hotspots are more likely to occur. The cases of coil burnout due to this have been reported, and such quenches are difficult to detect. Several methods have been proposed to detect and protect against quenches, but the coil temperature is already rising when a quench is detected. This means that coil operation must be stopped before the actual quench signal by predicting the occurrence of a quench. In this study, we show the results of quench prediction for unknown data by training data obtained from numerical simulations using a neural network called LSTM.