A PID algorithm based on convolutional neural network for pointing stabilization control of synchrotron radiation beams

A PID algorithm using a fixed set of parameters in synchrotron radiation beam control systems does not meet the need to control various environmental vibrations. This study proposes a vibration identification method based on convolutional neural network(CNN) to improve the PID algorithm. Firstly, the genetic algorithm is used to rectify the optimal control parameters for various types of vibration in advance, and then the CNN recognizes the vibration types to assist the PID algorithm in selecting the corresponding optimal control parameters online, so as to realize the optimal control under various vibrations. A set of mixed vibration signal inputs are applied to the beam control system to detect the final beam position deviation displacement. The experiment shows that the beam deviation displacement under the improved PID method is smaller and has better results than the traditional PID control method.