Adaptive PMLSM Control via Hybrid Extended State Observer Repetitive Control Optimized by Improved Deep Reinforcement Learning
Year: 2026
Type: article
Abstract: The active disturbance rejection controller (ADRC) is a widely used algorithm for suppressing speed and thrust fluctuations in permanent magnet linear synchronous motor (PMLSM) drives. However, traditional ADRC has limited disturbance rejection capability under multi-frequency harmonics, and its reliance on manual parameter tuning hinders adaptability to varying operating conditions. To address these challenges, this paper proposes an intelligent hybrid active disturbance rejection control (IHADRC) framework. The method integrates a hybrid extended state observer with repetitive control to simultaneously estimate and suppress both periodic thrust ripples and aperiodic disturbances. A grey wolf optimizer-optimized deep reinforcement learning framework enables intelligent parameter self tuning and enhances robustness and generalization. The HESO RC estimation error dynamics are proven uniformly ultimately bounded under bounded disturbance derivative and residual assumptions. Finally, the proposed IHADRC framework is implemented on a three-phase PMLSM platform and evaluated under 7 distinct operating conditions using 9 performance metrics. Experimental results show that the proposed IHADRC scheme reduces speed error RMSE by 40% and q-axis current tracking errors by 43% on average compared to HADRC, and by 57% and 54% respectively compared to PSOADRC, across 7 distinct operating conditions evaluated with 9 performance metrics.
Institution: Naval University of Engineering
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7166809198