Aiming at the uncertainty of hydrological model parameters and runoff simulation results, four objective functions, namely Nash-Sutcliffe efficiency coefficient (NSE), modified determination coefficient (R²_m), relative error (PBIAS) and Kling-Gupta efficiency coefficient (KGE), were selected to calibrate and verify the parameters of the constructed SWAT model at the upstream of Panjiakou Reservoir in the Luanhe River Basin. The sensitivity difference of model parameters under different objective functions and the uncertainty of runoff simulation were analyzed. The results show that the parameter sensitivity will change with the increase of the number of iterations and the change of the sampling range, and the parameter range and the optimal value of the calibration are significantly different under different objective functions. As the objective function, NSE and KGE are more robust in runoff simulation at each station, showing higher simulation accuracy and lower simulation uncertainty, respectively.