控制理论(社会学)
计算机科学
自适应控制
模型预测控制
控制系统
控制工程
控制(管理)
工程类
理论(学习稳定性)
扭矩
工作(物理)
病媒控制
噪音(视频)
作者
Jun Gu,宋士涛,Zhexi Xu,Dianguo Xu
出处
期刊:
日期:2026-09-01
卷期号:: 399-403
标识
DOI:10.1109/aees71194.2026.11745683
摘要
To address the degradation of speed-loop control performance caused by external disturbances and parameter uncertainties, this paper proposes a learning-enhanced predictive PI auto-tuning method for PMSM drives. A discrete-time speed-loop model and a finite-horizon cost function are constructed to balance tracking accuracy and control effort. The PI gains are updated online using a numerical-gradient-descent method combined with projection constraints, thereby ensuring bounded parameter adaptation. Furthermore, Lyapunov-based stability analysis demonstrates that, under bounded disturbances and finite-difference approximation errors, and with a sufficiently small learning rate, all closed-loop signals remain uniformly ultimately bounded. Simulation results show that, compared with conventional fixed-gain PI control, the proposed method achieves faster tracking response, stronger disturbance rejection, and improved robustness against inertia mismatch. The present validation is simulation-based; hardware implementation and comparison with additional adaptive baselines remain subjects of future work.
科研通智能强力驱动
Strongly Powered by AbleSci AI