Temperature-field-angle dependent critical current estimation of commercial second generation high temperature superconducting conductor using double hidden layer Bayesian regularized neural network

Abstract

Abstract We estimated the critical current of the second generation (2G) high-temperature superconducting (HTS) conductor using neural network fitting methods. The critical current of 2G HTS conductors depends on magnetic field strength and angle as well as on temperature, Ic(T, B, θ). Moreover, the critical current values vary for 2G HTS conductors from different manufacturers. In this study, we addressed three challenging issues in critical current assessment by neural network fitting methods, namely 90° asymmetry, a wide range of temperature-field-angle dependence, and different manufacturer conductor differences. Prediction models for three commercial HTS conductors were trained and evaluated by convergence, accuracy, and robustness. The linear regression correlation coefficient R was approximately equal to 1 for the three models. The critical current estimation obtained from the proposed method was compared with the critical current estimation from the interpolation method at different fixed temperatures using a multi-width no-insulation magnet. The model computation speed was also discussed. The proposed model needed only 2.7 s to compute 10 million data sets. Therefore, the convergence, accuracy, reliability, and speed of the proposed method prove that it can be used in a wide range of industrial applications and academic fields.