This paper presents a hybrid physics-data-driven optimization model for grassland grazing management. It comprehensively assesses essential factors in the Abaga Banner grassland ecosystem, including soil moisture, vegetation biomass, desertification degree index, and soil compaction. Through a thorough analysis, the impacts of grazing patterns and intensity on the grassland’s physical characteristics and biomass are studied. Employing genetic algorithms, an optimal grazing model is formulated to minimize soil desertification throughout the year. The paper aims to contribute to grassland ecology restoration and ensure sustainable livelihoods for local herdsmen, offering a scientific foundation for promoting the sustainable growth of grassland husbandry.