RESEARCH ON COOPERATIVE CONTROL OF MAGLEV TRAIN SUSPENSION SYSTEM BASED ON DEEP REINFORCEMENT LEARNING
-
-
Abstract
Due to the problem of multi-electromagnet coupling in the suspension frame of maglev train, the coupling effect will cause the unstable suspension of the suspension frame during operation. In order to ensure the stable operation of the multi-electromagnet in the suspension frame under external interference, a deep reinforcement learning collaborative control method for the suspension frame of electromagnetic levitation (EMS) type maglev train is proposed. Firstly, the dynamic modeling of multiple electromagnets in maglev suspension frame is carried out and their coupling characteristics are analyzed. Secondly, the multi-electromagnet cooperative control method (SAC-CC) based on SAC algorithm is proposed, and the collaborative control algorithm framework of deep reinforcement learning is constructed. The dynamic model of multi-electromagnet is converted into a deep reinforcement learning environment model, and the reward function is designed for this model. Then, the SAC-CC controller is obtained by training in static floating environment. Finally, the trained SAC-CC controller is applied to the suspension control and cooperative control of multiple electromagnets in the suspension frame under different working conditions. The effectiveness and robustness of the proposed controller are verified by comparing with the traditional proportional integral-differential (PID) control method. The results show that: Under different working conditions, compared with PID controller, the SAC-CC controller proposed in this paper can not only effectively control the multiple electromagnets stably suspended near the equilibrium point in the suspension frame, but also significantly reduce the coupling effect between electromagnets, and has better suspension control performance and collaborative control performance. The suspension control performance and cooperative control performance of SAC-CC controller under different working conditions are improved by about 30%-99% and 30%-75% respectively.
-
-