Reliable operation of standalone PV–battery DC microgrids is challenged by significant uncertainty in PV power generation, whichdegrades the performance of model predictive control (MPC) strategies—particularly with respect to DC bus voltage regulationand maximum power point tracking (MPPT). This paper presents a physics-informed offset-free MPC approach that explicitlycompensates for PV production prediction errors. The method incorporates real-time irradiance-model updates based on the PVphysical characteristics, enabling the controller to correct forecasting inaccuracies without requiring additional disturbance oroffset states. Simulation results show that conventional MPC becomes highly sensitive to irradiance prediction errors, causingdeviations from the maximum power point, reduced energy extraction and fluctuations in DC bus voltage. In contrast, the proposedcontroller maintains accurate MPPT operation, ensures stable DC-link voltage regulation and respects battery state-of-chargelimits even under large PV prediction uncertainties. Quantitatively, the method improves MPPT tracking accuracy and reducesDC-link voltage deviation compared with standard MPC across a wide range of operating conditions. These findings demonstratethat the proposed physics-informed offset-free MPC significantly enhances the robustness and reliability of standalone PV–batteryDC microgrids subject to uncertain solar conditions.
