Deep learning based surrogate models for first-principles global simulations of fusion plasmas

Abstract

Abstract The accurate identification and control of plasma instabilities is important for successful fusion experiments. First-principle simulations that can provide physics-based instability information such as the mode structure are generally not fast enough for real-time applications. In this work, a workflow has been presented to develop deep-learning based surrogate models for the first-principle simulations using the gyrokinetic toroidal code (GTC). The trained surrogate models of GTC (SGTC) can be used as physics-based fast instability simulators that run on the order of milliseconds, which fits the requirement of the real-time plasma control system. We demonstrate the feasibility of this workflow by first creating a big database from GTC systematic linear global electromagnetic simulations of the current-driven kink instabilities in DIII-D plasmas, and then developing SGTC linear internal kink instability simulators through supervised training. SGTC linear internal kink simulators demonstrate predictive capabilities for the mode instability properties including the growth rate and mode structure.