Artificial-Intelligence-Driven Model for Resistive Superconducting Fault Current Limiter in Future Electric Aircraft

Abstract

Fast and accurate electrothermal characterization of Superconducting Fault Current Limiters (SFCLs) is critically important for their performance evaluation. At the design stage of the SFCL, simulation is needed to validate the feasibility of the idea before fabrication. The most popular simulation models are the finite element models, thermal electrical analogy (TEA), equivalent circuit model and the finite difference method. However, all those four models are time consuming either in preparing stage or in running stage. In this study, we propose an Artificial Intelligence-driven model to achieve fast calculation and monitoring of SFCL. In this paper, a surrogate model was presented to estimate the resistance and temperature of SFCL, under different fault scenarios. These scenarios include different fault duration, voltages, maximum prospective fault current, and different operating temperatures. The TEA was used to electrothermally simulate the SFCL, under different fault scenarios. Then, temperatures of different layers in SFCL, as well as the total resistance of SFCL were collected. Afterwards, a Feed Forward Neural Network (FFNN) was used for SFCL characterization. The data of fault scenarios were fed into FFNN as inputs while temperature and resistance were considered as outputs. Also, to show the superiority of FFNNs, two other intelligent methods were used for surrogate modelling of SFCL. Results showed that FFNN had more than 98% accuracy for estimating different electrothermal parameters of SFCL. This intelligent surrogate model is a step forward towards developing digital twin for SFCL in future cryo-electric aircraft.

Intelligent surrogate model of a high-temperature superconducting cable for reliable energy delivery: short-circuit fault performance

“…Surrogate models use either mathematical equations or intelligent characteristics, try to find a correlation between inputs and outputs without directly solving the governing equations. In other terms, surrogate models could be considered as shortcuts through the complicated modelling [30][31][32][33].…”