Data-Driven Nonlinear Model Predictive Control for Power Sharing of Inverter-Based Resources

Model predictive control (MPC) is a closed-loop optimization framework that can solve the real-time control challenges of inverter-based distributed energy resources (DERs) in smart grids. This article addresses the challenge of heavy reliance of model predictive controllers on physics-based dynamic models by proposing a data-driven MPC framework via sparse regression (SR) theory and nonlinear model predictive control (NLMPC) framework. Unlike existing approaches that rely on approximate models based on physical principles or experiments, the proposed framework directly captures the dynamics of the DERs using measurements. This capability enables power sharing among DERs and active/reactive load support with high precision. The framework can capture uncertainties and drift dynamics of DERs by updating the data-driven model on a timely manner for running the MPC for effective power sharing. By employing this approach, the overall effectiveness of active and reactive power sharing is enhanced without compromising voltage and frequency control. The proposed optimal control strategy is validated through real-time simulations conducted on a 3-DER microgrid (MG) using OPAL-RT. The results demonstrate the successful estimation of DER dynamics using the SR method and accurate power sharing through NLMPC. Furthermore, NLMPC not only achieves a high degree of precision in power tracking but also outperforms other MPC strategies that rely on successive linearization, with a mean absolute percentage error (MAPE) of 6.83% for active power and 5.71% for reactive power.