Adjustable-Order Noise-Resilient Extended State Observer for Model-Free Predictive Speed Control

This article proposes a noise-resilient extended state observer (NR-ESO) with adjustable-order to enhance the noise suppression ability of model-free predictive speed control (MFPSC) for permanent magnet synchronous motor (PMSM) drives. Conventional MFPSC suffers from two main challenges: the inherent tradeoff between disturbance rejection and noise amplification in linear extended state observer (LESO), and performance degradation due to control mismatch between the nominal and actual control gain. To address these issues, an NR-ESO is developed by integrating an adjustable-order pre-filter. By enabling flexible pole placement for the augmented system, the proposed structure provides additional flexibility in shaping noise attenuation, thereby enhancing noise suppression without sacrificing dynamic performance. Furthermore, to deal with performance deterioration due to control gain uncertainty, an adaptive law is designed to estimate the control gain online, eliminating the dependence on mechanical parameters. Experimental results verify that, compared with the conventional methods, the proposed noise-resilient model-free predictive speed control (NR-MFPSC) reduces current ripples by over 70% while maintaining equivalent dynamic performance. The proposed method ensures strong robustness against mechanical parameter and load uncertainties, making it ideal for high-precision servo drives.