Model-Free Predictive Control of PMSMs Assisted by Modified Cascaded Extended State Observers

Abstract

This paper presents a model-free predictive control strategy for permanent magnet synchronous motor (PMSM) drives, aiming to enhance system robustness and dynamic performance under parameter variations and load disturbances. The proposed control scheme employs an ultra-local model to formulate a model-free predictive controller for the current loop, thereby eliminating reliance on accurate machine parameters. To achieve robust outer-loop speed regulation, a modified cascaded extended state observer (ESO) is integrated into the control structure. The cascaded ESO is designed to jointly estimate the total disturbance, while a disturbance decomposition mechanism is introduced to separately compute partially known components-such as viscous friction and load torque-and incorporate them into the ESO framework. This approach effectively reduces the observer estimation burden and improves disturbance reconstruction accuracy without increasing bandwidth or sensitivity to measurement noise. The combined control architecture enables real-time compensation of internal and external disturbances, ensuring precise current tracking and stable speed regulation. Comparative simulation studies demonstrate that, in contrast to conventional cascaded ESO-based methods, the proposed strategy significantly mitigates speed fluctuations, enhances robustness, and improves both transient response and steady-state accuracy.