To renovate an ordinary building roof into an extensive green roof (GR), on the basis of optimizing the runoff process, the total runoff control rate (TR-CR) and peak runoff control rate (PR-CR) were analyzed with the SWMM model under parameter sensitivity analysis. The response surface method (RSM) was used to investigate the effect of area ratio (GR area to total roof area), width (surface width per unit), and soil hydraulic conductivity on TR-CR and PR-CR. The optimal test conditions were obtained through the establishment of quadratic regression equations. The results showed that the effects of the three factors on TR-CR were in the order of area ratio>hydraulic conductivity>width, and on PR-CR was area ratio>width>hydraulic conductivity. Considering the volume capture ratio (80%) of annual rainfall and construction cost of the GR, under the optimal conditions of the area ratio of 0.79, width of 2.21 m, and hydraulic conductivity of 55.59 mm/h, the average values of TR-CR and PR-CR for this optimization scheme were 80.2% and 80.9%, respectively. This result show that the multiple quadratic regression model established is reliable. Therefore, the RSM can effectively optimize the simulation parameters, and provide some theoretical and practical basis for the renovation or construction of GRs.