Optimized reaction mechanism for hydrogen-ammonia combustion

Hydrogen and ammonia have attracted significant attention as promising carbon-free fuels for various combustion applications. As these fuels exhibit contrasting combustion characteristics, H2 being highly reactive and NH3 having relatively low reactivity, their mixture offers a way to balance these extremes. By blending H2 and NH3, it is possible to compensate for the shortcomings of each fuel and flexibly control the overall reactivity of the mixture, making it suitable for a wide range of practical applications. Although substantial progress has been made in the chemical kinetic modeling of H2–NH3 mixtures over the past decade, considerable uncertainties remain in numerical predictions. This project aimed to optimize the reaction kinetics of H2–NH3 mixtures to enhance the predictive performance of chemical reaction mechanisms under various conditions. A genetic algorithm–based optimization strategy was implemented using PyGAD to systematically refine the rate parameters of the most sensitive reactions in the kinetic mechanism developed by Alessandro Stagni et al. The optimization targets the minimization of discrepancies between experimental measurements and numerical simulations performed with Cantera, considering multiple observables, including ignition delay times and species concentration profiles across a range of thermochemical conditions. To ensure physically meaningful yet flexible parameter exploration, a data-driven statistical gene space was constructed following the REPRICA methodology. This approach enables probabilistic sampling of Arrhenius parameters beyond the restrictive analytical bounds traditionally imposed by uncertainty formulations, thereby expanding the feasible search domain while preserving consistency with prior kinetic knowledge. This work will contribute not only to the development of efficient combustion and energy systems based on H2–NH3 mixtures, but also to the advancement of chemical kinetic modeling for such fuel blends.