Abstract: This paper presents a geometric-probabilistic hybrid model for balancing computational efficiency and accuracy in calculating excavation volumes for pumped storage stations in reservoir perimeter expansion in the site selection phase. First, we construct a sector-weighted geometric excavation model that uses a unified reference excavation width to account for topographic constraints in various azimuthal directions, and solve it iteratively for the effective excavation volumes required to achieve the target reservoir capacity. Then, we adopt a one-dimensional lognormal distribution to characterize the elevation-dependent variability of weathered layer thickness, and locally weighted linear regression to estimate distribution parameters. And, analytical-numerical expressions are derived for the expected cross-sectional areas of distinct weathered layers, enabling a probabilistic estimation of stratified excavation volumes. In a case study of the Shulan pumped storage station, our new hybrid method and the recommended design scheme are compared with those of the conventional average thickness method. Results indicate that the overall error in total excavation volume calculations is reduced to lower than 20%, and it outperforms the latter significantly in estimating the excavation volumes for weathered layers. We have integrated this model into a code for optimizing the scale parameters of pumped storage stations, achieving an effective approach to scheme comparison in the site selection stage.

Key words: pumped storage power station, reservoir perimeter excavation, geometric-probabilistic hybrid model, locally weighted linear regression, lognormal distribution