Physics-model-based Real-time Optimization for the Development of Steady-state Scenarios at DIII-D

W.P. Wehner, A. Pajares, E. Schuster, J.R. Ferron, D.A. Humphreys, R.D. Johnson, B.G. Penaflor, K.E. Thome, M.L. Walker, C.T. Holcomb, B.S. Victor

27th IAEA Fusion Energy Conference

Gandhinagar, India, October 22-27, 2018

Abstract

In this work, a novel q profile control approach and recent DIII-D experimental results aimed at reaching stationary plasmas characterized by a flat loop voltage profile are presented. The control approach combines commands computed both offline (feedforward) and online (feedback). Both command components are computed via numerical optimal control techniques. The key advantage of the numerical computation approach is that it allows for the explicit incorporation of state and input constraints to prevent the controller from driving the plasma outside of stability limits and obtain, as closely as possible, stationary conditions characterized by a flat loop voltage profile. Using a suitable control-oriented model, the simulated plasma evolution in response to the actuators is embedded into a nonlinear optimization problem that provides a feedforward control policy (set of actuator waveforms) that under ideal conditions guides the plasma evolution to the desired state. The time trajectory of the plasma current, gyrotron power, and neutral beam power are optimized to guide the plasma to stationary state characterized by a flat loop voltage profile. It is shown in simulations that an overshoot in the plasma current during ramp-up combined with a particular timing of the gyrotron and neutral beam injection can improve the uniformity of the loop voltage profile. The feedback controller computes updates to the feedforward control law to account for variability in plasma conditions; optimizing in real-time the model-predicted plasma evolution in response to the available actuator set.