Advanced relaxed stochastic control for green energy management and decarbonization in large-scale heterogeneous industrial clusters

Abstract

Hydrogen is integral to decarbonizing high-temperature heat processes in heavy industries such as steel, glass, and aluminum, which collectively account for 10%–15% of global energy-related CO${}_2$ emissions. The increasing demand for industries based on hydrogen necessitates the development of advanced strategies for the management of hydrogen industrial clusters (HICs) driven by renewable energy sources. In this paper, a sophisticated controller is introduced to manage an HIC, considering uncertainties in thermal demand, energy forecasting, and energy and hydrogen prices. In order to ensure an entirely green energy system, the infrastructure integrates an electrolyzer, multiple compressors, multiple hydrogen storage tanks, and hydrogen-only gas burners. The HIC is designed to simultaneously support hydrogen injection into multiple hydrogen-dependent industries, including steel, glass, and aluminum. Moreover, it can operate in off-grid mode (without hydrogen market access) or on-grid mode (with hydrogen market access), optimizing resource utilization and energy management. In order to address uncertainties and reduce computational complexity, Boolean relaxations and the stochastic methodology are integrated into the model predictive control structure, and the main goals are unifying off- and on-grid operations and optimizing thermal demand fulfillment and tank management. Numerical simulations demonstrate that this strategy effectively manages multiple tanks in parallel configuration, ensuring the efficient HIC operation by fulfilling thermal demands, adhering to functional constraints, reducing costs, and enhancing revenue. Simultaneously, this approach leads to a 28% reduction in operational costs and a decrease of over 2200 switching events annually. It also enhances computational efficiency, achieving nearly 40% faster computation times using both open-source and commercial solvers, and in practice, this leads to significant improvements in overall system performance.