Actuator Management in Tokamaks via Receding-Horizon Optimization
A. Pajares, E. Schuster
47th European Physical Society (EPS) Conference on Plasma Physics (CPP)
Virtual Meeting, June 21-25, 2021
Abstract
|
|
For efficient and safe operation of reactor-grade tokamaks such as ITER,
one of the main control-design challenges is the optimal management of
the available actuators in order to fulfill the required control objectives.
This is a particularly difficult problem due to the multitude of control
tasks that share many of the device actuators. In addition, the appearance
of off-normal events and exceptions (due to changes in the plasma state
and/or actuator failures) normally makes off-line scenario planning
insufficient and requires real-time optimization solutions.
Along the lines of previous efforts towards tackling this problem [1],
this work proposes a novel actuator-management algorithm based on real-time
optimization (see Fig. 1). The individual controllers send to the
actuator manager high-level control requests, which are embedded into a
finite receding-horizon optimization problem. This is done by equating
the high-level control requests with their associated virtual-input functions,
which depend on the plasma state, actuator signals, and time. Whereas
the individual controllers compute so that they fulfill their specific
control goals, the main objective of the actuator manager is to find
the control input so that as many of the control requests as possible
are fulfilled. If all the constraints cannot be exactly fulfilled
simultaneously, then some of them are relaxed according to the relative
priorities assigned to the control requests. At the same time, physical
saturation limits are imposed on the control inputs, and a performance
metric is minimized over the prediction horizon. This metric depends on
the difference between the predicted and target states as well as on
the control effort. A key advantage of this scheme is that it allows
for an independent high-level design of the controllers while ensuring
their integration via optimal actuator allocation. The capabilities and
performance of the proposed scheme are tested in one-dimensional
simulations using COTSIM (Control Oriented Transport SIMulator).
[1] A. Pajares and E. Schuster, "Actuator Management via Real-time
Optimization for Integrated Control in Tokamaks,"
46th European Physical Society Conference on Plasma Physics (2019).