Study on two-layer rolling optimal scheduling of island micro-energy grid based on adaptive decision-making mechanism

To address the significant dynamic response disparities among heterogeneous energy sources in island micro-energy grids and the difficulty of conventional dispatching in reconciling real-time performance with economic efficiency, this paper proposes a bi-level rolling optimization dispatching strategy based on an adaptive decision-making mechanism. The strategy incorporates a seawater desalination-coupled hydrogen production subsystem into the micro-energy network architecture, establishing a five-dimensional coordinated supply system encompassing electricity, heat, gas, hydrogen, and freshwater. Within a model predictive control framework, intraday dispatching is decoupled into fast and slow control layers according to subsystem response speeds. An intelligent decision-making mechanism simultaneously addressing economic efficiency, stability, and scenario adaptability is embedded in the slow control layer: at each dispatching interval onset, real-time operating state deviations are quantified, a comprehensive performance function integrating stability and economic costs is formulated, and the uniqueness condition for the marginal equilibrium optimal threshold is derived. An online adaptive threshold calculation method based on predictive-horizon look-ahead evaluation is further proposed, making scenario adaptability an intrinsic property of the mechanism. This mechanism operates synergistically with the fast-slow hierarchical structure, which provides a response-characteristic-matched physical framework; together they achieve precise suppression of strong stochastic source–load fluctuations on islands. Simulation results show that, compared with conventional fixed-period unified optimization strategies, the dispatching frequency of slow-response equipment is reduced by 93.2%, the trajectory deviation rate decreases by 16.3%, and the total operating cost is lowered by 29.9%.