Overcurrent cycling refers to the procedure of imposing repetitive overcurrent to superconducting tapes/devices for characterizing their critical current reduction. Characterizing the overcurrent cycling behaviour of Rare Earth Barium Copper Oxide (ReBCO) tapes is a crucial step in the design process of High Temperature Superconducting (HTS) devices. Multiple overcurrent incidents during the operation of an HTS device can significantly decrease the total critical current, leading to potential quenches and failures. Data-driven models have been proposed in the literature to estimate the Critical Current Degradation Rate (CCDR) of ReBCO tapes under multiple overcurrent scenarios. However, these methods have exhibited notable errors in the range of 8%–11%, in the estimation of the critical current reduction. This paper proposed methods based on Artificial Intelligence (AI) techniques aimed at the challenges of conventional methods of CCDR estimation. Different AI-based techniques were proposed, tested, and compared to show the effectiveness of the proposed intelligent approach, including Support Vector Regression (SVR), Decision Tree (DT), Radial Basis Function (RBF), and Fuzzy Inference System (FIS). Experimental data on critical current values of ReBCO tapes subjected to multiple and repetitive overcurrent cycles were employed for this investigation. The results demonstrated that the Mean Relative Error (MRE) of the SVR method is 23%, for the DT model is approximately 0.61%, the MRE of the FIS model is well above 0.06%, and the MRE value for the RBF method is about 1.1 × 10−6%. Moreover, the proposed AI models offer fast test times, ranging from 1 to 11 ms. These findings highlighted the potential of using AI techniques to enhance the estimation accuracy of the risks associated with overcurrent events.
Read full abstract