The SCS-CN model is one of the most widely method for estimating surface runoff, which can quickly assess the effects of different ground covers on precipitation-runoff relationships. To explore the applicability of the Soil Conservation Service Curve Number Model under artificial grazing measures in the Loess Plateau region, to improve the understanding of the impact of vegetation restoration on water resources in the Loess Plateau, and to provide a theoretical reference for the estimation of surface runoff from artificial grazing in the Loess Plateau, this paper selects data from 73 rainfall runoff events in seven runoff plots at the Luoyugou test site of the Tianshui Soil and Water Conservation Experiment Station of the Yellow River Water Conservancy Commission, and uses the step-size method to estimate the runoff from artificial grazing in the Loess Plateau. The runoff curve number (CN) and the initial abstraction ratios (λ) of the standard SCS-CN model were optimized by the step method. The results showed that: (1) The optimal initial abstraction ratios value for artificial grazing was 0.15, while that for control farmland was 0.18; (2) The difference of curve number values for different species of pasture was not significant, about 79, and that for control farmland was 81, indicating that the curve number values recommended by the U.S. Soil Conservation Service were applicable to the calculation of artificial pasture runoff on the Loess Plateau, but the the initial abstraction ratios values needed to be optimized and rate determined according to the actual situation; (3) The S values were modified according to the previous soil moisture S value was corrected according to the previous soil moisture condition, and the simulation effect was more satisfactory, and the modified SCS-CN model was applied to the semi-arid artificial grazing area of Loess Plateau.