Considering process efficiency, flow and coking are important aspects in biomass and its derivatives for catalytic conversion to value-added products. However, these investigations are scarcely involved due to the very complex reaction environment and coupling mass transfer and coking process. To better understanding the coking process in glucose catalytic conversion to levulinic acid (LA), a particle-scale model based on the lattice Boltzmann model (LBM) was proposed. In this model, a coking model based on to volume of pixels scheme was developed to simulate the coking deactivation process, and a zeolite permeability model based on the gray LBM scheme was developed for fluid permeation inside shaped zeolite. The model was validated by both analytical and experimental data. The effects of temperature, particle porosity, and coking on catalyst deactivation, product yield, and selectivity were investigated. The numerical results indicated that the deactivation time of shaped zeolite peaked at the particle porosity of 0.20 and increased as the operating temperature decreased. Glucose conversion, LA yield, and humins yield increased with the increase of particle porosity and operating temperature. The opposite trend of the intermediate 5-hydroxymethylfurfural (HMF) was found when the temperature and porosity were increased. The results also showed that the LA selectivity and the yield ratio of LA to humins peaked at a porosity of 0.45 and a reaction temperature of 180 degrees C.