Design of microgrids with ammonia-based energy storage via Bayesian optimization

As renewables-based microgrids face challenges of intermittent and uncertain renewable energy sources, ammonia offers a promising solution for energy storage due to its high energy density and low storage cost. The design of microgrid systems has to account for the highly time-sensitive availability of renewable resources, where both long-term trends and short-term uncertainties must be considered. This work presents an optimization framework that addresses the design of microgrids with ammonia-based energy storage under uncertain wind speed profiles, where Bayesian optimization is employed to efficiently navigate the design search while incorporating a rolling-horizon scheduling framework that allows the evaluation of the operating costs for different scenarios. Computational case studies investigate the influence of connectivity with the main grid, showing that islanded operation requires significantly larger storage and production units for hydrogen and ammonia. The proposed method found microgrid designs that clearly outperform those obtained from random search, the particle swarm optimization algorithm, and a deterministic combined design and scheduling approach. A sensitivity analysis on the flexibility of the Haber-Bosch process also demonstrates that extending its operating range from 90%–100% of its capacity to 10%–100% yields a 18% cost reduction, underscoring the importance of flexible ammonia production.