Speaker
Description
AC losses significantly impact the design and operation of superconducting power devices, making their fast and accurate calculation essential. However, calculating AC losses in high-temperature superconducting coils often requires substantial computational resources. This paper proposes an artificial neural network (ANN) model with a Transformer architecture to calculate AC losses in high-temperature superconducting coils. The Transformer model is integrated to capture long-range dependencies and complex interactions between input features such as frequency, temperature, and current density, which significantly influence AC losses. The self-attention mechanism of the Transformer allows the model to focus on the most influential parts of the input data, thereby improving prediction accuracy. The ANN is trained using data derived from finite-element analysis (FEA) to accurately estimate AC losses. The model minimizes output error by adjusting the weights and thresholds. The response set Y at sample points X is obtained through FEA calculations, and an AC loss surrogate model is constructed using the dataset of sample points X and their corresponding responses Y. The accuracy of this surrogate model is validated by comparing its predictions with those obtained from COMSOL simulations. The results demonstrate that the Transformer-enhanced surrogate model can significantly improve computational efficiency without sacrificing accuracy.
