Abstract
This study presents a novel methodology using machine learning (ML) and 3D GIS building models to assess solar irradiance on building envelopes in complex urban and mountain environments. Four ML models were developed to predict hourly global horizontal solar irradiance (GHI), ensemble modelling and stacking techniques were employed for their combination. The predicted GHI values were then used with the Perez model to estimate hourly global tilted solar irradiance (GTI) on the 3D GIS building model. The results demonstrate high accuracy, with a strong correlation between predicted and measured GHI (R2: 0.904-0.973) and precise simulation of GTI on building envelopes (R2: 0.814-0.924). This approach contributes to more effective and efficient methods for predicting solar irradiance in challenging environments.