Abstract: Scientific and rational modes of grazing utilization and fertilization management are important agronomic measures to maintain the production performance and community structure stability of legume-grass mixtures. The study aimed to explore the regulation and control mechanisms of grazing and fertilization on the production performance and community structure stability of perennial legume-grass mixtures, and to preliminarily determine the optimal mode of grazing utilization and fertilization management for legume-grass mixtures. In the Hexi Corridor area, a 4-year-old mixed grassland with Medicago sativa ‘Qingshui’, Bromus inermis, and Elytrigia elongata (at 1∶1∶1) were used as the study objects. The L₁₆ (3¹×4²) mixed level orthogonal experimental design was used with three heights prior to mowing: 20 cm (S₁), 30 cm (S₂), and 40 cm (S₃); four mowing intensities: 2 cm (E₁), 5 cm (E₂), 8 cm (E₃), and 11 cm (E₄); four nitrogen application rates: 0 kg(N)∙hm⁻² (N₁), 75 kg(N)∙hm⁻² (N₂), 150 kg(N)∙hm⁻² (N₃), and 225 kg(N)∙hm⁻² (N₄); and the impacts of artificial mowing to simulate grazing utilization combined with nitrogen application on the community structure and production performance of mixed grassland were monitored and analyzed. The study showed that the height prior to mowing and the amount of nitrogen added had a greater impact on the production performance of mixed grasslands. The total biomass of grassland under the S₃E₂N₄ treatment was the highest (17 707.80 kg∙hm⁻²), the crude protein content of mixed forage under the S₁E₁N₄ treatment was the highest (15.46%), and the relative feeding value of mixed forage under the S₃E₁N₃ treatment was the highest (184.93). Additionally, reducing the mowing intensity could increase utilization times and the proportion of gramineous forages in the community. Among them, the coverage ratios of leguminous forages and Gramineae grasses under the treatments S₂E₃N₄, S₁E₃N₃, S₁E₄N₁, S₂E₄N₃, and S₁E₁N₁ were close to 1∶1 after simulated grazing utilization, and the treatments S₂E₃N₄, S₃E₃N₁ and S₃E₄N₂ had the best effect on the number of forage branches. Furthermore, compared with the hay harvested by conventional mowing (cutting three times a year in the area), simulated grazing reduced the forage yield but greatly improved the nutritional value of the forage and could effectively inhibit the excessive growth of M. sativa ‘Qingshui’ in the community. At the same time, it maintained the stability of the legume-grass mixture community structure. In conclusion, when the management measures were as follows: the height prior to mowing was 30 cm, mowing intensity was 8 cm, and nitrogen application rate was 225 kg(N)∙hm⁻² (S₂E₃N₄), the stability of community structure and composition balance of mixed grassland were the best. In addition, mixed grassland also had good production performance (total biomass of grassland reached 15 173.41 kg∙hm⁻²; crude protein content of mixed forage reached 13.92%; relative feed value of mixed forage reached 156.93) under the S₂E₃N₄ treatments. This management measure is a suitable reference for grazing utilization and fertilization management of mixed grasslands composed of M. sativa, B. inermis, and E. elongata, and may be popularized and applied in the Hexi Corridor and other similar regions. It can provide a scientific basis and technical support for optimizing the production and utilization of local legume-grass mixtures.