AN ADAPTIVE NEURAL PI CONTROL FRAMEWORK USING DEEP REINFORCEMENT LEARNING FOR HIGH-PERFORMANCE PMSM SPEED DRIVES
DOI:
https://doi.org/10.62985/j.huit_ojs.vol26.no2E.400Từ khóa:
Permanent magnet synchronous motor, adaptive PI control, deep reinforcement learning, DDPG, neural networksTóm tắt
High-performance PMSM speed control remains challenging due to parameter uncertainties, nonlinear dynamics, and load disturbances. Although conventional PI controllers are widely adopted, their fixed gains limit adaptability under varying operating conditions. To overcome this, this paper proposes an adaptive DRL-based neural PI control framework for PMSM drives. The method integrates a DDPG algorithm with an RBF neural network for continuous online gain tuning. Furthermore, an anti-windup mechanism and a reference model–based augmentation are incorporated to ensure closed-loop stability under actuator saturation. The learning objective is formulated to optimize tracking accuracy, transient performance, and disturbance rejection. Simulation results demonstrate that the proposed framework significantly outperforms fixed-gain PI controllers, minimizing settling time, overshoot, and steady-state error under ideal physical constraints. These findings confirm the framework's robustness and efficiency for advanced motor drives.
Tài liệu tham khảo
[1] B. K. Bose, Modern Power Electronics and AC Drives, Upper Saddle River, NJ, USA: Prentice Hall, 2002.
[2] R. Krishnan, Permanent Magnet Synchronous and Brushless DC Motors, Boca Raton, FL, USA: CRC Press, 2010, doi: https://doi.org/10.1201/9781420014235
[3] P. Vas, Vector Control of AC Machines, Oxford, U.K.: Oxford Univ. Press, 1990.
[4] K. J. Åström and T. Hägglund, Advanced PID Control, Research Triangle Park, NC, USA: ISA-Instrumentation, Systems, and Automation Society, 2006.
[5] M. S. Boucherit, “Adaptive speed control of permanent magnet synchronous motor,” in Proc. Int. Conf. Electrical, Electronic and Computer Engineering (ICEEC’04), 2004, pp. 1–6, doi: https://doi.org/10.1109/ICEEC.2004.1374633
[6] O. Dogru, K. Velswamy, F. Ibrahim, Y. Wu, A. Sundaramoorthy, B. Huang, S. Xu, M. Nixon, and N. H. Bell, “Reinforcement Learning Approach to Autonomous PID Tuning”, in Proceedings of the 2022 American Control Conference (ACC), Atlanta, GA, USA, 2022, pp. 2691–2696, doi: https://doi.org/10.23919/ACC53348.2022.9867687
[7] W. J. Shipman and L. Coetzee, “Reinforcement learning and deep neural networks for PI controller tuning,” in Proc. 18th IFAC Symp. Control, Optimization and Automation in Mining, Mineral and Metal Processing (MMM 2019), 2019, pp. 1–6, doi: https://doi.org/10.1016/j.ifacol.2019.09.173
[8] P. Lu, J. Xiao, F. Zhou, and W. Hu, “Adaptive proportional–integral robust control of an uncertain robotic manipulator based on deep deterministic policy gradient,” Guangzhou Univ., Guangzhou, China, Tech. Rep., 2021, doi: https://doi.org/10.3390/math9172055
[9] W. Xu, J. Wang, D. Liu, et al., “A deep reinforcement learning-based control method for electric linear loading systems” J Mech Sci Technol 39, 7815–7827, 2025. Doi: https://doi.org/10.1007/s12206-025-1139-8
[10] T. Haarnoja, A. Zhou, P. Abbeel, and S. Levine, “Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor,” arXiv:1801.01290, 2018, doi: https://doi.org/10.48550/arXiv.1801.01290
[11] A. Kumar et al., “Deep learning-enabled adaptive control of electric drives,” IEEE Trans. Power Electron., Apr. 2023, vol. 38, no. 4, pp. 5000–5012, doi: https://doi.org/10.1109/ACCESS.2025.3648165
[12] W Cui, Y Jiang, B Zhang, “Reinforcement learning for optimal primary frequency control: A Lyapunov approach,”IEEE Transactions on Power Systems, 2023, vol. 38, no. 2, pp. 1676–1688, doi: https://doi.org/10.1109/TPWRS.2022.3176525
[13] J. Yao, M. Han, and X. Yin, “Lyapunov-based distributed reinforcement learning control with stability guarantee,” Computers & Chemical Engineering, Apr. 2025, vol. 195, Art. no. 108979, doi: https://doi.org/10.1016/j.compchemeng.2024.108979
[14] F. Li and Y. Du, Eds., Deep Learning for Power System Applications. Cham, Switzerland: Springer, 2024, doi: https://doi.org/10.1007/978-3-031-45357-1
[15] M. J. Hossain and S. Debnath, “Neural-Network-Based Intelligent Control for Permanent Magnet Motors: A 2025 Perspective,” Journal of Control and Automation, vol. 15, no. 2, pp. 112-128, Feb. 2025, doi: https://doi.org/10.1038/s41598-025-02396-y
