COTSIM-based Gain Optimizer for Feedback Control of Advanced Scenarios in NSTX-U
Y. Manjetta, H. Al Khawaldeh, T. Rafiq, E. Schuster
Division of Plasma Physics (DPP) Annual Meeting of the American Physical Society (APS)
Pittsburgh, PA, USA (Remote), November 8-12, 2021
An optimizer has been developed to tune the gains of fixed-structure
scenario controllers in NSTX-U. As indicated by their name, fixed-structure
controllers have a well-defined structure but are functions of
to-be-tuned gains. This is the case, for instance, of the widely used
Proportional-Integral-Derivative (PID) controller, which is usually
empirically tuned. As alternative, a more systematic approach to gain
tuning is proposed in this work by selecting the gains based on a
predefined, user-defined, optimization criterion. This criterion is
expressed in terms of a to-be-minimized cost function weighing different
metrics associated with closed-loop control performance such as tracking
overshoot/undershoot, rise times, disturbance rejection, and steady-state
tracking error. This optimization approach is enabled by closed-loop
simulations carried out by using the Control Oriented Transport SIMulator
(COTSIM) for NSTX-U, which is in turn wrapped around an optimizer based
on the Sequential Quadratic Programming (SQP) technique. SQP is capable
of solving arbitrary nonlinear optimization problems while handling
constraints on the to-be-optimized parameters. The effectiveness of the
method is illustrated for several scenario-control objectives.
*Supported by the US DOE under DE-SC0021385.