Automatic Modeling of High-Temperature Superconducting Cable Using Reinforcement Learning

Abstract

This paper proposes a technique of automatic modeling for high-temperature superconducting (HTS) cables. Reinforcement learning (RL), which is a representative methodology for automation and intelligence, is applied for the automation of the proposed modeling. To reflect the high-frequency characteristics of the HTS cables in the proposed modeling, reflectometry-based cable modeling is used. In addition, for agent training, an environment that combines simulation and experiment results is proposed, and detailed techniques for the process of the proposed RL model are introduced. The proposed technique is demonstrated by experiment using an actual HTS cable under 77 K and 300 K conditions. It is expected that the proposed technique will allow anyone without the related knowledge to perform the modeling of the HTS cables.