Yiying Wang, Chen Liu, Xuedong Li, Yanhui Wu, Ying Gu, and Fei Wang.
Permanent magnet synchronous motor, vector control, deep reinforcement learning, SAC, current tracking
This paper presents an advanced intelligent control strategy aimed at reducing the tracking error in permanent magnet synchronous motor (PMSM) by employing the soft actor–critic (SAC) algorithm within a deep reinforcement learning (DRL) framework. The current control problem is modelled as a Markov decision process (MDP), and a physical model of the PMSM vector control system is developed in the MATLAB Simulink environment. The SAC controller is incorporated into the current loop, enabling dynamic interaction and iterative learning with the motor environment. Simulation results, compared with proportional–integral (PI), DDPG, and TD3 controllers, demonstrate that the SAC significantly enhances the tracking accuracy of the d–q axis current under three typical operating conditions. Moreover, the SAC improves speed control accuracy and exhibits superior robustness to load disturbances. These findings highlight the potential of SAC to effectively improve both the performance and stability of motor control, offering substantial prospects for precision motor control applications.
Important Links:
Go Back
