We describe the offline machine learning (ML) development for an effort to precisely regulate the Gradient Magnet Power Supply (GMPS) at the Fermilab Booster accelerator complex via a Field-Programmable Gate Array (FPGA). As part of this effort, we created a digital twin of the Booster-GMPS control system by training a Long Short-Term Memory (LSTM) to capture its full dynamics. We outline the path we took to carefully validate our digital twin before deploying it as a reinforcement learning (RL) environment. Additionally, we demonstrate the use of a Deep Q-Network (DQN) policy model with the capability to regulate the GMPS against realistic time-varying perturbations.
Cite
@article{arxiv.2105.12847,
title = {Developing Robust Digital Twins and Reinforcement Learning for Accelerator Control Systems at the Fermilab Booster},
author = {D. Kafkes and M. Schram},
journal= {arXiv preprint arXiv:2105.12847},
year = {2021}
}
Comments
Corresponding proceedings for poster presentation at 12th International Particle Accelerator Conference - IPAC'21; final version