A surrogate modeling approach for evaluating the shading effect on building energy performance

This study presents a novel surrogate modeling approach to predict indoor thermal environments in dense urban contexts. By explicitly incorporating key shading parameters—average surface Sky View Factor (SVF) and sunlight hours (SH)—the model addresses limitations in conventional surrogates that overlook or simplify surrounding configurations. Indoor air temperature was selected as the primary output metric to directly capture thermal responses to urban geometry without the confounding effects of building systems. Validation results show high accuracy (MAPE: 1.25%, MAE: 0.215°C). Sensitivity analysis confirms that excluding SVF or SH significantly degrades predictive performance (MAPE increases of 8.87% and 6.86%, respectively). In fixed urban contexts, core zone volume becomes the dominant factor, while west-facing zones show highest sensitivity to shading effects—revealing how variable importance shifts across different urban configurations. These findings underscore the critical role of SVF and SH in capturing the shading effects essential for accurate indoor temperature prediction. • This study develops a surrogate model for predicting energy with shading effects. • We introduce a model using SVF, sunlight hours, weather data, and building geometry. • SVF and sunlight hours effectively capture shading impacts on energy performance. • Sensitivity analysis identifies key parameters for energy prediction accuracy.