Dynamic building performance gap related to occupant behaviors: A framework to predict the impacts of possible causal factors

• Predicts building performance gaps from occupant behavior parameters. • Neuro-fuzzy prediction model improves accuracy of energy simulations. • Stochastic inputs analyzed across five building typologies. • Up to 16.24 % better energy prediction vs. calibrated EnergyPlus model. • Framework supports resilient nZEB performance forecasting. This paper aims to develop a prediction framework for minimizing Building Performance Gaps (BPGs) considering the frequent dynamics in non-residential buildings by quantifying the impacts of occupant behavior (OB) parameters. The framework is designed as an extension to existing building performance simulation tools, enabling the quantification of OB discrepancies between design-stage predictions and operational outcomes. Through detailed simulation analyses across eight standard building models, we examine the influence of three independent OB parameters: occupancy schedules, occupancy activities, and setpoint temperatures. Among these, setpoint temperature is identified as the main OB parameter in determining energy demand based on duration curve of different standard models. While the occupancy schedules and activities have negligible impact on peak power demand, they contribute significantly to the energy performance gap (up to 50 %) and deviations in indoor air temperature (up to 5 %). We develop and train the neuro-fuzzy prediction model using the results of simulation-based analysis to estimate the effects of each OB parameters in BPGs. The prediction framework is validated using a calibrated case study, where results demonstrate an improvement of 16.24 % in energy performance prediction and 5 % in indoor temperature estimation compared to standard simulations. While the framework effectively increases simulation accuracy in predicting building frequency energy performances, it is currently limited to simulation-based evaluations.