A Unified Novel Koopman-Based Model Predictive Control Scheme to Achieve Seamless Stabilization of Nonlinear Dynamic Transitions in Inverter-Based Stochastic Microgrid Clusters

The pursuit of seamless formation and robust control of inverters in power electronic-dominated grids face challenges arising from uncertainties in grid impedance, dynamic load variations, and transitional phase jump scenarios, leading to elevated instability during mode transitions. These challenges necessitate a nuanced approach to voltage regulation and stability analysis to manage the stochastic nonlinearities and inherent dynamics effectively. This paper introduces a novel Unified Koopman-based Model Predictive Control (K-MPC) method that synergizes a modified MPC framework with an ensemble approach and a saturation-like automatic adaptation function for seamless inverter transitions. It facilitates precise power-sharing among inverters in isolated microgrids by adjusting output impedances without the need for communication lines, thereby addressing the limitations in dynamic response, sensor requirements, filter fluctuations, and controller complexity. The K-MPC method enhances system performance and fidelity by incorporating online adaptation norms for external disturbance rejection, aligning closely with a predefined reference model. Quantitative validation, conducted through frequency response analysis and hardware-in-the-loop (HIL) experiments on an IEEE 123-node test system, underscores the method's effectiveness. The K-MPC approach notably reduces total harmonic distortion (THD) by up to 30% relative to conventional control strategies and improves power-sharing precision among inverters by 25% under dynamic loading conditions. Furthermore, it exhibits a 40% faster response in adapting to external disturbances, ensuring voltage and frequency remain within target thresholds.