Actuator Management via Nonlinear Real-time Optimization
A. Pajares, E. Schuster, M. Walker, D. Humphreys
Division of Plasma Physics (DPP) Annual Meeting of the American Physical Society (APS)
Remote, November 9-13, 2020
Abstract
|
|
An actuator management algorithm based on nonlinear, real-time
optimization techniques has been designed, tested in simulations, and
is currently being implemented within the DIII-D Plasma Control System
(PCS). Tokamaks are highly complex devices in which a multitude of
control tasks must be carried out by many shared actuators. This
motivates the development of actuator management algorithms within
present and future PCS designs. The primary goal of an actuator manager
is to calculate in real time actuator requests that can fulfill as many
control objectives as possible despite physical saturation limits and
potential actuator failures. In this work, an actuator management
algorithm has been designed based on an Augmented Lagrangian optimization
method. Different control objectives can be embedded in this scheme
both as terms within a cost function that needs to be minimized (e.g.,
control effort) and as constraints (e.g., absolute or rate limits for
power, torque, etc.). Its performance has been tested in DIII-D simulations
using the Control-Oriented Transport Simulator (COTSIM). Moreover, it
is currently being implemented within the DIII-D PCS as a further step
towards developing a higher level of integration for the present
architecture.