Potential and challenges of urban building surface solar energy utilization in solar resource non-enriched areas, China

The assessment of solar utilization potential faces three key challenges: vast building inventories, computationally intensive high-fidelity radiation modeling, and integrating multidimensional utilization indicators. To address these, this study develops an integrated urban solar assessment framework that combines building cluster segmentation with four technical advancements: temporal-spectral sky modeling, automated facade orientation detection, high-performance coupled computing, and intelligent metric analytics. Systematic evaluation of machine learning algorithms identified eXtreme Gradient Boosting (XGB) as the most effective model (R2 > 0.95, MSE <0.10) for predicting solar potential in non-enriched urban areas. The findings indicate that (1) Urban building cluster variability is primarily driven by geographic, morphological, and typological parameters essential for high-accuracy solar potential modeling. (2) Seasonal radiation disparities exhibit significant latitudinal dependence, with winter irradiance ranging from 30 % to 63 % of summer levels across studied cities, while shading effects obstruct 34.70–50.71 % of facades compared to just 3.45–6.98 % of roofs. (3) Solar utilization potential follows a nonlinear threshold response to solar accessibility, necessitating optimized, orientation-specific harvesting strategies.