A new control strategy addresses a critical challenge in decarbonized energy systems: coordinating variable renewable generation with hydrogen production serving multiple applications. Researchers developed a mixed-integer Model Predictive Control (MPC) framework for microgrids that integrate solar photovoltaics, wind turbines, and proton exchange membrane electrolysers, designed to supply both continuous heating demands and discrete hydrogen refueling station requests.
The framework's innovation lies in handling two fundamentally different hydrogen demand patterns. Heating applications require steady hydrogen supply, while mobility refueling involves intermittent, scheduled deliveries via tube-trailer dispatch. Traditional control methods struggle with such temporal mismatches. The researchers introduced a critical-time mechanism that ensures timely fuel dispatch while avoiding unnecessary grid electricity imports when renewable output falls short.
The study compares fixed and shrinking prediction horizons over 12-hour and 24-hour windows—a distinction rarely explored in prior hydrogen microgrid research. Results demonstrate that shrinking horizons cut computational complexity by up to 70% while maintaining nearly identical performance. Renewable utilization exceeded 96% for heating applications and 90% for mobility, with hydrogen production and grid consumption varying by less than 0.1% across different horizon configurations.
These findings have immediate implications for system operators. Computational efficiency matters because electrolysers operate at scales requiring real-time optimization across dozens of constraints. The ability to reduce processing time while preserving renewable integration rates makes MPC practical for deployment in actual microgrids rather than simulation only.
The work also validates a design principle: shrinking horizons, which adapt prediction windows as time advances, outperform fixed windows in managing operational complexity. This scalability is essential as hydrogen-integrated grids proliferate. System planners can now confidently implement such controllers knowing they will maintain high renewable penetration while serving diverse end-use demands without excessive computational burden.
