Abstract: SWAT model is widely used in the field of water science, and the SUFI-2 algorithm is a preferred tool for SWAT optimization. It is extremely important to improve the calibration efficiency of the SUFI-2 algorithm for SWAT application. Three objective functions, NS, R2 and PBIAS, were set up in the SUFI-2 algorithm. Then, the sensitivity of 14 hydrological parameters was analyzed, and the range of parameters was calibrated based on the limitation principle of 95PPU’s uncertainty (P-factor>0.7 and R-factor <1), so the best simulation was selected according to the classification criteria of simulation performance evaluation. Finally, the simulation results were verified based on the hydrological data of Chushandian Reservoir from 1980 to 2018. The results showed that among the three objective functions, PBIAS has the strongest response to parameter change. The sensitive parameters determined by PBIAS were CN2, CANMX, SOL_BD, GW_DELAY, GWQMN, GW_REVAP, ESCO, RCHRG_DP and SOL_AWC. The simulation uncertainty in the calibration period and validation period was related to the precipitation distribution. The best simulation performed stably and efficiently in the two periods, which proved that the selection method was feasible. The research results can be used as reference to improve the optimization efficiency of hydrological models.