Optimization of a methanol/NOx combustion mechanism based on a large amount of experimental data

Investigating the methanol (CH3OH) / nitrogen-oxides (NOx) combustion system is an important task since methanol is a promising alternative to fossil fuels, and its interactions with nitrogen oxides are significant due to environmental effects. The performances of the recently available detailed mechanisms in simulating the experimental results are still unsatisfactory. The aim of this work is to develop a more reliable reaction mechanism using parameter optimization. First, the Glarborg-2018 mechanism was updated with the rate parameters of the previously optimized H2/NOx and methanol mechanisms of ELTE. A large collection of literature data was compiled, which consists of direct measurements and theoretical determinations of the rate coefficients (2175 data points in 130 data series), indirect measurements of the formaldehyde (CH2O) /NOx and CH3OH/NOx system in homogenous reactors (2373 data points in 225 data series), and the neat CH3OH and CH2O subsystems in homogenous reactors and flames (689 data points in 68 data series). Using code Optima++, we optimized the rate parameters of the 24 most important elementary reactions that were identified by the recent PCALIN (Principal Component Analysis of the Parameter-Uncertainty and Data-Uncertainty Scaled Local Sensitivity Matrix with Linear Corrections) active parameter selection method as most influential. The optimized rate coefficients were assessed in detail and compared with literature data. The optimized mechanism can reproduce the CH2O/NOx and CH3OH/NOx combustion experimental data on average within their 3.5σ experimental uncertainty, which means it performs significantly better than the previously published mechanisms, which have average errors larger than 5σ. The reproduction of neat CH3OH and CH2O experimental data also improved. The optimized mechanism was also tested on experimental data of the H2, H2/NOx, syngas, and syngas/NOx combustion systems. In all cases, the optimized mechanism reproduced these experimental data better than the initial mechanism, although these data were not used as optimization targets.