A Deep Learning-Based Ultra-Fast Surrogate Model for AC Loss Estimation in Superconductors Using COMSOL

High-temperature superconductors (HTS) demonstrate the potential to significantly benefit the energy and transport sectors by enabling high-current densities in cryogenic environments with minimal losses and exceptional efficiency. Finite element methods are widely used to model the electromagnetic behavior of HTS but are computationally intensive and time-consuming. The artificial intelligence (AI) method and its related techniques such as machine learning are advanced techniques to be used for the modelling of HTS in different large-scale power applications. AI-driven models are accurate and can be used for real-time simulation, together with correlating complex relations between inputs and outputs in short computation time in a fraction within one second and less. In this paper, we propose the use of built-in surrogate models in COMSOL Multiphysics and develop a deep learning surrogate model based on simulation data natively in COMSOL to estimate AC losses in HTS tapes. The training data was generated using the 2D FEM model using <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">H</i> formulation. This approach improves the speed of prediction to almost immediate prediction and maintains a high level of accuracy with a final root mean square error (RMSE) of 0.9990. In addition, the results of the built-in surrogate model have been compared with other AI-based models. The potential of using surrogate modelling to accurately predict AC losses for future superconducting power applications is demonstrated and discussed by analyzing the deep learning model and results.