Optimization of LOHC systems using machine-learning-based surrogate models for integrated low-carbon hydrogen and power supply in cement production

Industry decarbonization is a priority to mitigate climate change. Cement production is a hard-to-abate sector responsible for 7% of emissions, for which hydrogen has been proposed to reduce energy-related emissions. However, storage remains a critical challenge as conventional approaches require extreme operating conditions. LOHCs may overcome this limitation by allowing ambient storage compatible with existing fossil infrastructure. This work presents a novel integration of LOHCs with a hybrid CHP (SOFC and hydrogen turbine) in the cement process, considering both technical and performance indicators. Two configurations were studied to supply: (a) hydrogen only, and (b) hydrogen and power. Exhaust heat from CHP and cement process was proposed to meet the dehydrogenation thermal demand. LOHCs dehydrogenation and CHP were modeled using first-principles models. A multiobjective optimization was formulated to minimize CAPEX while maximizing LOHCs utilization efficiency. This problem included continuous and discrete variables, resulting in a mixed-integer nonlinear programming (MINLP) problem. Given the high computational cost, a surrogate-based optimization strategy was adopted. Common regression and classification models were trained on a dataset generated via maximin Latin Hypercube sampling. Reactor operating conditions, LOHCs feed and catalyst use were the decision variables. The final design reached for Cases (a) and (b) 53.82% and 50.66% LOHCs utilization efficiency, defined as the fraction of LOHCs-stored hydrogen converted into power and fuel for the cement process, with CAPEX of 1.58 and 3.37 MM$/MW respectively, covering the dehydrogenation process and CHP module. Integrating LOHCs with cement production and CHP reduced dehydrogenation energy losses. This positions LOHCs as a promising route toward cement sector decarbonization.