Multi-input multi-output model predictive heat flux controller based on a time-dependent scrape-off layer simulation

Burning plasma devices need to be prepared for the high amount of power exhaust that escapes the core region and reaches the divertor plates. Current methods of developing controllers for such devices are limited as they require system identification run time in uncontrolled scenarios. Here we propose the use of time-dependent scrape-off layer simulations such as SOLPS-ITER to identify the system dynamics of heat flux and its response to multiple impurity seeding in the form of Neon gas seeding (cooling peak ~ 30-40 eV) and Boron powder dropping (~ 3-5 eV). We have developed an IMAS data model-compatible synthetic bolometer that emulates real bolometer output based on time-dependent plasma state information from simulations. With the identified system model and realistic gas injection and powder dropper models with associated latency and non-linearities, we have developed a multi-input multi-output model predictive controller. The multiple inputs allow robust wide phase space operation while multiple impurity outputs allow for efficient cooling at different temperatures. The predictive capability of the controller counteracts the latency and directional behavior of actuators. This controller has been optimized and tested with closed-loop simulations without the use of device operating time. This controller development workflow can be utilized to design an initial controller for a future burning plasma device as well as to help design bolometer line of sight placement and actuator design.