Molecular-Level Kinetic Modeling of Biomass Gasification

A molecular-level kinetic model for biomass gasification was developed and tuned to experimental data from the literature. The development was divided into two categories: the composition of the feedstock and the construction of the reaction network. The composition model of biomass was divided into three submodels for cellulose, hemicellulose, and lignin. Cellulose and hemicellulose compositions were modeled as linear polymers using Flory-Stockmayer statistics to represent the polymer size distribution. The composition of lignin, a cross-linked polymer, was modeled using relative amounts of structural building blocks or attributes. When constructing the full biomass composition model, the fractions of cellulose, hemicellulose, and lignin were optimized using literature-reported ultimate analyses. The reaction network model for biomass contained pyrolysis, gasification, and light-gas reactions. For cellulose and hemicellulose, the initial depolymerization was described using Flory-Stockmayer statistics. The derived monomers from cellulose and hemicellulose were subjected to a full pyrolysis and gasification network. The pyrolysis reactions included both reactions to decrease the molecule size, such as thermal cracking, and char formation reactions, such as Diels-Alder addition. Gasification reactions included incomplete combustion and steam reforming. For lignin, reactions occurred between attributes and included both pyrolysis and gasification reactions. The light-gas reactions included water-gas shift, partial oxidation of methane, oxidation of carbon monoxide, steam reforming of methane, and dry reforming of methane. The final reaction network included 1356 reactions and 357 species. The performance of the kinetic model was examined using literature data that spanned six different biomass samples and had gas compositions as primary results. Three data sets from different biomass samples were used for parameter tuning, and parity plot results showed good agreement between the model and data (y(predicted) = y(obs)0.928 + 0.0003). The predictive ability of the model was probed using three additional data sets. Again, the parity plot showed agreement between the model and experimental results (y(predicted) = y(obs) 0.989-0.007).