Quantifying uncertainty in multi-scale energy analysis of residential archetypes: A stochastic occupancy approach

• Developed a Probabilistic Occupancy Model (POM) for multi-scale residential energy analysis. • Trained POM with Time Use Survey (TUS) data to predict four occupancy states with uncertainty. • Achieved 90 % accuracy for nighttime and 64 % for daytime occupancy predictions. • Validated POM for multiple spatial and temporal scales energy consumption estimation. • Statistical tests confirmed POM’s superiority over deterministic and traditional occupancy models. Occupancy models are key input parameters in Building Energy Modelling (BEM). Typically, multi- scale residential energy analyses use conventional occupancy schedules which overlook their inherent uncertainty and stochasticity. This oversight can lead to inaccurate energy consumption predictions, as building-level occupancy models fail to capture urban-scale occupancy dynamics. This study formulates a robust and widely applicable methodology that creates Probabilistic Occupancy Model (POM) for residential dwellings based on Time Use Survey (TUS) data which are applicable across any geographical area or region. The multi-class classification model predicts four occupancy states (NotActive, Outside Activity (OA), Light Activity (LA), Medium Activity (MA)). The POM achieved 90 % accuracy in predicting nighttime occupancy states and an average accuracy of 64 % during the day, reflecting variations in daytime activities. The effectiveness of the POM is evaluated through a multi-scale energy analysis of Irish residential archetypes, demonstrating high accuracy across spatial scales. On average, the model predicted annual heating energy consumption within 0.5 % of measured values across national and regional scales. For quarterly and monthly energy consumption, the POM achieved accuracy within 1 %, and for daily predictions, the model’s estimates closely aligned with measured peaks and troughs, showing an average error below 2 %. The findings underscore the value of probabilistic occupancy modeling in improving energy consumption estimates, demonstrating POM’s accuracy across spatial scales and its potential to enhance urban energy planning and efficiency strategies.