In Europe, the energy renovation of the existing building stock is a great opportunity to significantly reduce energy consumption and greenhouse gas (GHG) emissions and reach the European sustainability targets. In this framework, building energy simulations (BES) tools are very useful in verifying energy retrofit measures’ effectiveness and compliance with national standards. However, an inaccurate numerical prediction, the so-called “performance gap” between measured and numerical performance, is often obtained, mainly due to the inherent uncertainty of model input. Due to its stochastic nature, the occupants’ behavior (OB) is considered among the key contributors to this gap. However, the most recent Building Energy Model (BEM) approaches adopt deterministic hourly-defined profiles for characterizing OB, thus neglecting the related uncertainty. In this work, the impact of OB uncertainties on energy consumption (EC) prediction is evaluated by adopting a Karhunen-Loève Expansion sampling technique, used to randomly perturb OB profiles such as heating setpoint (HS), internal thermal loads (IL), and windows opening (NV). Two BEMs of a typical Italian residential building in pre- and post-renovation scenarios are considered and calibrated on real EC data. The results demonstrated that HS uncertainty has the highest impact on EC in all scenarios. Moreover, the higher the energy performance of the building, the higher the impact of OB, especially for IL and NV patterns.
