Invariant Manifold Based Model-Free Predictive Control for PMSM Motor with Adaptive Time Constant

Most existing finite control set model-free predictive control (FCS-MFPC) methods based on ultra-local model involve complex function structures and multiple hyperparameters, which inevitably increase the complexity of system structure. To overcome these limitations, this paper proposes an observer-less FCS-MFPC scheme suitable for permanent magnet synchronous motor (PMSM) based on invariant manifold constraint. The key lies in utilizing a low-pass filter to construct constraints, thereby defining a stable convergence subspace for the online computation of unknown nonlinear terms. In this way, only one parameter, namely the time constant needs to be tuned. Meanwhile, to balance the estimation accuracy and the robustness, the time constant of the low-pass filter is online optimized using the gradient descent method for different operating scenarios. The proposed method is analyzed using Lyapunov theory to ensure system convergence. The proposed method presents a simple structure with high robustness. Finally, experimental results validate the practical effectiveness of the proposed method. Specifically, it demonstrates a 50 - 70% reduction in both steady-state torque ripple and phase current THD compared to the traditional method, alongside enhanced robustness against parameter mismatch.