JACoW is a publisher in Geneva, Switzerland that publishes the proceedings of accelerator conferences held around the world by an international collaboration of editors.
TY - UNPB AU - Jiao, Y. AU - Wan, J. ED - Schaa, Volker RW ED - Huang, Wenhui ED - Yan, Xueqing ED - Tang, Chuanxiang ED - Li, Lu TI - Machine learning application studies in lattice design and optimization of 4th generation synchrotron radiation light source J2 - presented at SAP2023, Xichang, China, 10-12 July 2023 CY - Xichang, China T2 - Symposium on Accelerator Physics T3 - 14 LA - english AB - For the 4th generation synchrotron radiation light source, it is required to reach ultralow emittance in the lattice design, while keeping satisfactory nonlinear performance. In the design stage of a low-emittance storage ring, the beam dynamics can be very complicated and cause great challenges to optimization, due to strong chromaticity compensation sextupoles used in the storage ring. In this talk, we report several machine learning (ML) applications that speed up the design and optimization process of a ultralow emittance storage ring. Two representative works will be introduced. We developed a neural network-based genetic algorithm that can significant improve the convergence rate and diversity among solutions compared to conventional genetic algorithms, and we trained a ML model to learn the correlations between the short-term tracking results and the long-term motion stability of particles travelling in the ring, which can reduce the computing cost of the dynamic aperture by at least an order of magnitude. ER -