Speaker
Description
In modern synchrotron light sources, maintaining beam stability is critical for ensuring high-quality synchrotron radiation performance. Light source stability is governed by stability of current, beam position and beam size. Beam size stability on the order of several microns need to be improved for future experiments. Reinforcement learning (RL) offers a promising approach for real-time beam size feedback system. The RL framework consists of an intelligent agent that interacts with the environment to maximize a cumulative reward, based on state observations and actions. Beam size measurement at one point and vertical dispersion are observations of RL environments, which can present beam size distribution along the storage ring. Through simulation and real experimental setups, we demonstrate the efficacy of the PPO algorithm which adapt to discrete action spaces in controlling beam stability and correcting coupling. Further optimization of hyperparameters during simulation environment is applied in real operation. The approach provides a significant improvement in online, real-time correction of coupling errors, offering a faster and more adaptable solution compared to traditional methods.
