Research on beam dynamics optimization of a storage ring based on machine learning

This paper discusses the machine learning methods for the beam dynamics optimization in a storage ring for synchrotron light sources. A new method of closed orbit feedback based on machine learning was developed to enhance synchrotron radiation stability, which was then piloted at Shanghai Synchrotron Radiation Facility (SSRF). We use a novel machine learning technique to calibrate the linear optics, which differs from the traditional singular value decomposition (SVD)-based linear optics from closed orbit (LOCO). Predicting dynamic apertures is also one of the machine learning applications in our work.