Input–Output Variation-Driven Gain Observer for Ultra-Local-Model-Based Predictive Control of PMSMs Without Dependence on Electrical Parameters

Model-free predictive control (MFPC) based on ultra-local model (ULM) is characterized by enhanced robustness to parameter mismatches. However, the selection of control gains, actually, still relies on the parameters, and the gain deviation degrades the control performance. To solve this issue, this paper proposes an improved MFPC with a novel input-output-variation driven gain observer. With a conceptual ULM-based MFPC, the impacts of gain deviation on the control performance and system stability are analyzed in detail. Then, the proposed method is expatiated, and in particular the stability of the proposed gain observer is proven based on the Lyapunov stability theorem. Meanwhile, the design guideline of feedback coefficients of the proposed observer is also presented. Based on the theoretical analyses, a finite-control-set model free predictive current control (FCS-MFPCC) method with the suppression of gain deviation is designed and applied to a permanent magnet synchronous machine (PMSM) drive system. The digital delay compensation and the discriminator of zero voltage vectors are considered for performance improvement and inverter switching frequency reduction, respectively. Finally, some critical experiments are conducted based on a PMSM test rig to verify the proposed gain observer and FCS MFPCC method. In particular, different initial values of control gains are set, indicating that the proposed FCS-MFPCC method does not require any prior knowledge of electrical parameters of the PMSM under test.