Triple Validation of Calibrated Building Energy Models with Different Air Infiltration Values

Model calibration refines design-stage inputs to align with real-world building performance. Accurate parameter selection, especially for highly sensitive variables like air leakage, is crucial. This study compared two building energy model calibration methods. The “classic” method adjusted indoor air capacitance, internal mass, and air infiltration, while a novel method focused on capacitance and internal mass, using empirical data for infiltration. The infiltration values were calculated using the decay equation and the EnergyPlus equations with site-specific coefficients. A triple validation assessed model performance in terms of temperature (CIBSE TM63), energy consumption (minimization), and indoor air quality (represented by ${\mathrm{CO}}_2$ levels in accordance with the ASTM D5157 Standard). Results demonstrated the novel method’s superiority across all three performance metrics. All calibrated models met the CIBSE TM63 criteria even during the validation period, which was five times longer than the training period. Compared to the classic method, models incorporating dynamic empirical infiltration showed a 29% and 26% improvement in MAE and RMSE, respectively, in temperature prediction. In energy consumption results, the novel method models presented a 31% reduction, and for ${\mathrm{CO}}_2$ level agreement, these models achieved a 130% higher R² value than the classic model. In addition, the classic method’s infiltration values failed to meet ASTM D5157 requirements, suggesting reliance on unrealistic parameter values for accurate temperature representation. The incorporation of calculated air leakage data into the BEM allowed a more realistic estimation of capacitance and internal mass values, emphasizing the importance of accurate air infiltration modeling for overall model reliability.