In this work, a mechanistic model was developed to simulate the kinetics of the production of levulinic acid (LA) from sugarcane bagasse (SCB), rice husk (RH) and soybean straw (SS). The production of LA from those agro-industrial wastes followed the methodology of biorefining in three stages. Experimental data from the third stage (catalytic depolymerization of cellulose) obtained under a wide range of operating conditions were used to estimate the parameters of the model. An optimization procedure based on a genetic algorithm was used to determine the optimal parameter values. The prediction of the concentrations of glucose, 5-HMF and LA using the mechanistic model was particularly accurate, as demonstrated by R2 and RMSE. Thereby, a satisfactory concordance was reached between the high yields of LA of 61.1 mol%, 67.7 mol% and 61.4 mol% calculated by the model, and the experimental yields of 60.5 ± 2.1 mol%, 65.2 ± 2.9 mol% and 61.5 ± 4.0 mol% (under optimum conditions of 190 °C, 7.0% w/v of H2SO4, 75 min) for SCB, RH and SS, respectively. The biorefining of the agro-industrial wastes under optimum operating conditions allowed a satisfactory catalytic depolymerization of cellulose, regardless of the degree of crystallinity. The estimation of yields led to suggesting a strategy from the point of view of process synthesis and design, integrating SCB, RH and SS to supply their off-season and, thus, to ensure the supply of raw materials in the production of LA Ver menos