Abstract: Controlling the stability of surrounding rock during the excavation of large underground powerhouses represents a key challenge for the safe and efficient construction of pumped storage power stations. Taking the underground powerhouse project of the Daya River Hydropower Station (Daya River HPS) as the study object, this research integrated field hydraulic fracturing measurements and multivariate linear regression inversion to establish the initial in-situ stress field model for the powerhouse area, laying the foundation for subsequent excavation scheme analysis. For three proposed excavation sequence schemes, the entropy weight-TOPSIS method was applied to assign weights to indicators across three critical aspects: principal stresses, displacements, and plastic zone distribution. This enabled an objective evaluation of surrounding rock stability. Scheme I demonstrated a relative closeness coefficient of 0.78, proving superior to the other two schemes and was thus selected as the optimal excavation plan. The study demonstrates that the evaluation results align with the mechanical response laws of the surrounding rock revealed by numerical simulation. The findings provide a basis for subsequent support design and safe construction, while also offering significant theoretical reference and a practical case study for the design of excavation schemes in similar complex geological conditions.