Solar energy utilization on building surfaces in high-density cities

Reliable city-scale assessment of building-surface solar potential is essential for BIPV planning in high-density cities, yet existing methods often underrepresent facades, interrupt shading continuity, and rely on static screening criteria. This study develops a physics-data hybrid framework to evaluate solar energy utilization on full building surfaces in Shenzhen. The framework integrates multisource GIS data, dynamic building-cluster partitioning with preserved cross-cluster shading, Perez-based multidimensional sky models, high-resolution surface discretization, dynamic solar energy utilization thresholds, multi-scenario energy-benefit assessment, and an interpretable XGB surrogate model. Results show that rooftops receive 203.16 TWh of actual annual solar energy, while facades receive 176.38 TWh, confirming that facade resources are too significant to neglect in dense cities. The effective utilization ratio is strongly negatively correlated with the shading ratio (Pearson’s r = −0.97). In photovoltaic electricity generation, rooftops contribute 61.55% and facades contribute 38.45%. Under relaxed dynamic thresholds, the effective utilization ratio of north facades increases from 0.19 to 0.88, indicating that refined thresholds reveal substantial latent deployable potential and improve seasonal supply-demand evaluation. Under the theoretical full-surface deployment scenario and the assumed installation and efficiency settings, BIPV generation could cover 108.68% of Shenzhen’s building operational electricity demand. The XGB model achieves strong predictive performance (R² = 0.957–0.993). SHAP analysis identifies radiation type, dynamic threshold, and temporal setting as the primary drivers of model outputs, while morphological variables show secondary but context-dependent effects. Overall, the proposed framework provides a scalable tool for differentiated BIPV deployment and low-carbon urban energy planning in high-density cities.