Multi-Step Constrained Continuous-Control-Set Model Predictive Direct Power Control for Single-Phase Grid-Connected Inverters With Full-Order Lf-Cf-Lg Dynamics
Abstract
This paper presents a constrained model predictive direct power control (MPDPC) framework for a single-phase grid-following (GFL) inverter. The proposed method replaces separate outer power and inner current PI controllers with one predictive layer that directly tracks active and reactive power references. The constrained model predictive control (MPC) retains the current-control priority for GFL service while avoids intermediate current-reference generation. The online nonlinear MPC formulation optimizes the converter voltage over a prediction horizon while explicit state and input constraints are included. Simulation results verify the accurate active and reactive power tracking. Furthermore, parameter uncertainty tests under grid-inductance variation $\Delta L_{g} \in[-50 \%,+200 \%]$ confirm robustness without re-tuning. Compared with a conventional PI-based cascaded controller, the proposed unified MPC reduces rise time by up to 82.28%, settling time by 84.04%, and root-mean-squared error (RMSE) by 2.99%, while the PI baseline becomes unstable at $\Delta L_{g}=+200 \%$.