In the process of production of levulinic acid by acid catalytic hydrolysis of cane molasses,taking the yield of levulinic acid as the main indicator, single factor and combined effects of reaction time, the concentration of molasses, molasses-catalyst mass ratio , temperature were evaluated based on single-factor experiments. Box-Benhnken center combination experimental principle and response surface methodology(RSM) were adopted to optimize the preparation conditions of levulinic acid. A quadratic polynomial regression model was established to predict the LA yield. The results showed that the regression of this model was significant while the lack of fit was not significant, which indicated the model fitted well with experimental data. Each factor alone could significantly influence the yield of LA, and the order was molasses-catalyst mass ratio > temperature >effects of reaction time>the concentration of molasses. The highest LA yield was obtained under the optimal condition of 40 wt % molasses, 23.40:1 molasses - H2SO4 mass ratio at 151℃ for 5 h. The average yield of LA reached 30.11 wt% under the optimum conditions for three times. Compared with the predicted value of 30.96 wt%, the relative deviation was only 0.85 wt% , which proved the model was reasonable and feasible.