Towards Reactor-Relevant Control Strategies with RTCAKENN in Diagnostically Challenged Environments
2024 Research Campaign, Plasma Control
Purpose of Experiment
The purpose of the experiment is to progress towards the development of economically viable fusion power plants (FPPs) by demonstrating the capability to predict and control plasma profiles in real-time, even in the absence of diagnostics. This includes upgrading and testing RTCAKENN, a machine learning-based system, to provide real-time accurate profile estimations and control in diagnostically challenged environments. The experiment aims to enhance control over plasma profiles such as ion temperature and rotation in the absence of key diagnostics like CER (Charge Exchange Recombination) and TS (Thomson Scattering).
Experimental Approach
The experiment is structured into several stages to assess and optimize the LRAN and RTCAKENN controllers for electron temperature (Te) and rotation profile control:
Stage 1: Establish a reference shot using a super stable H-mode to serve as a baseline for the controllers’ performance.
Stage 2: Test LRAN for Te profile control, with and without inputs from diagnostics like CER and TS. This stage will involve 7 to 9 shots.
Stage 3: Focus on single profile control of rotation using LRAN, with a similar approach to Te control, involving 7 to 9 shots.
Stage 4: Test the RL controller for Te profile control under full and reduced RTCAKENN input configurations, using 2 to 4 shots.
The overall strategy is to adjust the experiment based on real-time results, conducting a total of 17 to 23 shots for sufficient testing and optimization.