Generating a simplified reaction model for methane and natural gas combustion using a genetic algorithm

Abstract

A methodology is presented to develop compact, high-fidelity simplified reaction models for hydrocarbon combustion using virtual species and simplified reaction pathways, with rate parameters optimized via a genetic algorithm (GA). The method was applied to methane and natural gas combustion, targeting key combustion properties: ignition delay times (IDT) and laminar burning velocities (LBV). The approach combines a detailed H₂/CO core with virtual reactions representing the main fuel oxidation pathways through fuel, fuel radical, and aldehyde virtual species. For natural gas, fuel components were lumped, and averaged thermodynamic properties were assigned to the virtual species. The optimization process produced simplified models with 14 species and 57 reactions, which could accurately reproduce the IDT and LBV simulation results of the AramcoMech 3.0 detailed model across a wide range of equivalence ratios and temperatures. The mean absolute deviations for all test conditions were 11.9% for IDT and 2.5% for LBV in methane, and 10.5% for IDT and 1.4% for LBV in natural gas simulations. The models could capture the tendency differences between methane/air and natural gas/air mixtures in ignition characteristics while preserving the similarities in flame propagation. The proposed method offers a practical alternative to conventional reduction techniques, enabling the generation of simple yet accurate reaction models suitable for CFD simulations in practical combustors with significantly reduced computational cost.