A Fast and Adaptive LSTM-based Surrogate Model for Predicting Limitation Performance of SFCLs in Electric Aircraft Systems

Superconducting fault current limiter (SFCL) is a promising device for improving the reliability, safety, and stability of electric aircraft by quickly and effectively limiting fault currents. There is a need to build a surrogate model that can fast and accurately predict the dynamic output of SFCL performance for aircraft system level studies as well as for developing digital twins. In this work, an electrothermal coupled finite element model of the SFCL was established to obtain the dataset during different fault scenarios. Then, a surrogate model is constructed using a long short-term memory (LSTM) network to predict the temperature, current, voltage, and resistance dynamic behaviors of SFCLs under faults. The results show that the proposed LSTM model can quickly predict the dynamic waveforms of SFCLs with different tape and fault parameters. The accuracy of prediction of the LSTM surrogate model is above 97%, and the relative error of the waveform predictions is lower than 5%.