Optimizing energy flow in advanced microgrids: a prediction-independent two-stage hybrid system approach

Abstract This paper presents a two-stage optimization framework for long-term energy management in microgrids, aiming to efficiently integrate various energy sources, storage systems, and consumption elements while addressing uncertainties in load demand and renewable generation. The framework consists of an offline optimization stage and an online optimization stage, each with distinct roles to balance long-term planning and real-time adaptability. In the offline stage, a robust two-stage mixed-integer linear programming (MILP) model is used to set annual targets for the state of charge (SoC) of energy storage systems. This stage applies a min-max-min approach to optimize for worst-case scenarios, establishing a cost-effective and reliable baseline plan that reduces dependency on conventional power sources and minimizes load deficits. The online stage, on the other hand, employs a new online convex optimization model that dynamically adjusts energy storage and dispatch decisions based on real-time data, allowing the microgrid to respond flexibly to fluctuations in demand and renewable generation. Simulation results using the Elia and North China datasets demonstrate the effectiveness of this two-stage approach. Offline optimization achieved up to 25% cost savings and reduced unmet demand by up to 99%, providing a stable foundation for efficient energy management. The online optimization stage further improved system responsiveness, minimizing reliance on backup generators and enhancing load reliability. This combined framework offers a comprehensive solution for optimizing microgrid performance, balancing predictive planning with real-time adaptability in complex, variable energy environments.