A Revisit to Model-Free Control

Abstract

Starting from Fliess&#x0027;s model-free control (MFC) technique developed 15 years ago, this paper aims to provide a systematic framework for characterizing, benchmarking, and generalizing this emerging control technique, with a particular focus on power electronics (PE). It examines the performance of MFC in terms of dynamic response, stability, and robustness, using the classical control theory as a basic tool. A theoretical comparison is conducted with the conventional linear control techniques on dynamic response and performance robustness. A generalized MFC theory and means to enhance its robustness performance are also highlighted. This paper suggests that MFC, in contrast to the conventional understanding based on model-independent error dynamics, is practically a model-based control technique. Such model dependency characteristics under MFC become more severe for PE systems. However, following a new design principle, MFC is found possible to possess extraordinarily robust performance against model variations as compared to most existing model-based control methods. On top of that, stability margin is found to be the key bottleneck hindering the performance robustness of the existing MFC techniques. A new MFC with greater stability margin and performance robustness is proposed in the paper. Comprehensive Pareto fronts analysis, simulations, and experiments are conducted on a buck converter system to verify the new understandings and conclusions drawn from the framework. With the new design principle and the new MFC, the system is able to demonstrate an almost constant dynamic response despite 25-fold circuit parameter (R, C, and L) variations and 1.85-fold input voltage variations.<xref ref-type="fn" rid="fn1">1</xref> <fn id="fn1"><label>1.</label><p>Manuscript received April 18, 2022; revised July 3, 2022; accepted Aug 4, 2022. This work was supported by the Australian Research Council under Grant DP220100231. Corresponding author: Sinan Li. W. Li, S. Li, and J. Zhu are with the University of Sydney, Australia (e-mail: wanrong.li&#x0040;sydney.edu.au; e-mail: sinan.li&#x0040;sydney.edu.au; e-mail: jianguo.zhu&#x0040;sydney.edu.au). H. Yuan is with Nanyang Technological University, Singapore (e-mail: huawei.yuan&#x0040;ntu.edu.sg).</p></fn>