Towards Scenario Optimization and Control in NSTX-U by Machine-Learning-Enhanced Equilibrium and Transport Modeling Using COTSIM

B. Leard, T. Rafiq, E. Schuster

33rd Symposium on Fusion Technology (SOFT)

Dublin, Ireland, September 22-27, 2024

The Control Oriented Transport SIMulator (COTSIM) is a dual equilibrium and transport code that captures the most significant dynamics of a tokamak discharge with calculation times suitable for control purposes. The modular nature of the code facilitates multipurpose functionality by allowing the user to choose models suitable to their necessary accuracy/speed requirements. Therefore, the code can range from fast to faster-than-real-time computational speeds depending on the objectives of the user. This work presents simulation results from COTSIM for several NSTX-U scenarios and compares them with TRANSP predictions and experimental measurements. The prediction capabilities of COTSIM for NSTX-U have been recently improved by including various neural-network surrogate models as well as a self-consistent equilibrium calculations. The sources’ depositions are now predicted using surrogate models for NUBEAM [1] (for neutral beam injection) and TORIC [2] (for heating harmonic fast waves). Similarly, a surrogate model for the Multi-Mode Module (MMM) [3] has been developed to predict the anomalous thermal, momentum, and particle diffusivities which are pivotal to modeling the evolution of temperature and rotation. Each of these surrogate models was trained specifically for the NSTX-U operating regime, with the goal of improving the accuracy of COTSIM’s predictions without significantly sacrificing computational speed. In addition, various fixed-boundary and free-boundary equilibrium solvers have been coupled with the transport solvers in COTSIM to predict self-consistent evolutions of the equilibrium and the plasma profiles throughout the tokamak discharge. These substantial modeling improvements allow COTSIM to be utilized for a wider range of model-based control objectives related to NSTX-U such as dual equilibrium and transport scenario optimization, real-time profile (e.g., temperature and rotation) estimation from limited/noisy measurements, and feedback-based advanced-scenario control.

[1] M.D. Boyer et al., Nuclear Fusion, 59 056008 (2019).
[2] Á. Sánchez-Villar et al., EPS Conf. Plasma Phys. 47A: o5.104 (2023)
[3] T. Rafiq et al., Phys. Plasmas 20, 032506 (2013).

*Supported by the US DOE under DE-SC0021385.