Taking the mass fraction of n-propanol as the control index and the lowest heat load as the objective function, the main parameters of double-column pressure-swing distillation process were determined by single factor optimization with the UNIQUAC thermodynamic model using PRO/II process simulation software. On this basis, with the theoretical number of trays, feed position and reflux ratio as independent variables, and the mass fraction of n-propanol as the response value, a second-order regression model was established through Box-Benhnken central combination design and global optimization of process parameters was carried out. The correlation parameters show that the regression model can be used to analyze and predict the relationship between response values and independent variables. Under the conditions of optimum process parameters, the absolute deviation between the experimental value of n-propanol mass fraction and the predicted value by the model is 6.86%, which  indicate  that the second-order regression model is reliable and effective.