Faster-than-real-time predictions and feedback control algorithm design and testing enabled by TRANSP

M. D. Boyer, K. G. Erickson, J-H. Ahn, J. Carlsson, S. Kaye, R. Nazikian, E. Schuster, S. A. Sabbagh

KSTAR Conference

Muju, South Korea, February 21-23, 2018

Abstract

Present-day and next step tokamaks will require precise control of plasma conditions, including the spatial distribution of rotation and current profiles, in order to optimize performance and avoid physics and operational constraints. This motivates expanding the availability of diagnostics in real-time as well as developing physics-model-based approaches to real-time plasma condition estimation, feedback control, and scenario forecasting. This work describes several ways in which the interpretive and predictive modeling code TRANSP is being used to enable model-based control development. TRANSP analysis has been used to develop reduced models capable of faster-than-real-time execution for forecasting and optimization, including a neural network that enables rapid evaluation of the beam heating, torque, and current drive profiles. TRANSP has also been used to identify control-oriented models for use in profile control algorithm design on NSTX-U. Finally, a method for implementing feedback algorithms within high-fidelity TRANSP simulations has been used to verify controller performance with the goal of reducing experimental commissioning time. A more general approach to control algorithm testing that enables linking TRANSP to Simulink is currently being developed. To support the increased real-time computing resources required for real-time predictive modeling, new hardware and software solutions to real-time communication in the Plasma Control System are being studied. Plans for applying the approaches described here on KSTAR will be discussed.