A BIM-Integrated Adaptive Surrogate Framework for Efficient Uncertainty Quantification in Building Energy Simulation

Abstract: Uncertainty analysis is critical for reliable building energy predictions, yet classical Monte Carlo simulation is computationally prohibitive for detailed physics-based models. This study proposes a BIM-integrated adaptive surrogate uncertainty analysis framework combining adaptive sampling with sparse Polynomial Chaos Expansion to efficiently propagate uncertainty and perform global sensitivity analysis. The approach is demonstrated on four Irish residential archetypes with uncertain envelope thermal parameters. The adaptive surrogates achieve stable accuracy (CV-RMSE 0.74–2.05 kWh/m² for annual EUI) while reducing computational cost by around 99% compared to Monte Carlo simulation, enabling scalable and interpretable decision support for building design and retrofit.