Incorporating Gas Puffing Delays into Density Control Synthesis in Tokamak Reactors By Combining Optimal Control and Reinforcement Learning Techniques
S.T. Paruchuri, E. Schuster
65th Division of Plasma Physics (DPP) Annual Meeting of the American Physical Society (APS)
Denver, CO, USA, October 30 – November 3, 2023
Gas puffing is one of the main actuation systems available for regulating
particle density in tokamaks. In next-generation tokamaks like ITER, the
distance between the gas valves and the vacuum vessel is expected to
delay the response of gas-puffing actuation systems. If not appropriately
accounted for during control synthesis, such delayed response from the
actuators can lead to performance degradation or destabilization of the
density control system. This work presents a systematic approach to designing
a model-based density controller that inherently accounts for input delays
in the gas puffing systems. The proposed control synthesis relies on
formulating an infinite-horizon optimal control problem and solving it
using reinforcement learning algorithms with an actor-critic architecture.
The resulting solution is then reformulated to account for input delays.
In addition to control synthesis, systematic methods to handle changes
in plasma operating conditions, which can lead to substantial uncertainties
in the density model, are also proposed in this work. Finally, the effectiveness
of the proposed density controller is illustrated using nonlinear numerical
simulations.
*Supported by the US DOE (DE-SC0010661, DE-SC0021385).