Occupants’ behavior and its impact upon the energy performance of buildings

Buildings account for up to 40% of the primary energy consumption in Europe. Therefore, ambitious energy saving targets of the European Union (Energy Roadmap 2050) have to include building stock energy retrofit. However, many field test studies show a discrepancy between expected and measured energy consumption for both new and retrofitted buildings. Reasons for this discrepancy are engineering system issues and occupants behavior. The aim of this work is to evaluate and model the impact of occupants’ behavior on the energy performance of buildings, with a particular focus on natural ventilation. The evaluation and modeling activity is based on high time resolution monitoring data collected from 90 apartments located in nine buildings retrofitted with different technologies. The expected heating energy consumption of the demonstration buildings is calculated following the EnEV procedure (German buildings’ energy saving ordinance). The expected and the observed heating energy consumption are then compared, and the degree of success of each retrofit layout is evaluated. The impact of occupants’ behavior is further analyzed through dynamic simulations. The heating energy consumption of the buildings is calculated based on two simulation scenarios: The first scenario considers constant ventilation rates, while the second scenario includes the ventilation losses due to the natural ventilation activities of the occupants. Thus, the observed state of the windows is provided to the simulation models as an input: this permits the simulation of the air change through each window. The first benefit of including the observed window state profiles in building energy performance simulation is a more precise prediction of the building energy performance; a second benefit is the inclusion of the heterogeneity of the occupants’ behavior in the simulation results. The collected data is then used to model the behavior of occupants, related to the change of state of the windows’. Two classes of occupants’ behavior models are therefore proposed. The first class of models is based on the first-orderMarkov chain technique with time-dependent transition probability matrices: Three different models for the generation of window state profiles are introduced. Hence, the models are validated through the Markov chain Monte Carlo technique. Once the profiles have been generated, they can be used as an input for dynamic building energy performance simulation purposes. The dynamic simulation of two demonstration buildings with the generated window state profiles shows reliable building energy performance predictions. The second class of models is based on the logistic regression analysis. The probability of action is inferred depending upon the time of the day, the weather conditions (temperature, humidity and wind speed) and the indoor air conditions (temperature, humidity and air quality). Within this methodology, the drivers for occupants to change the state of the windows are identified: Main drivers for the opening action of windows are the time of the day and the indoor carbon dioxide concentration. Main drivers for the closing action of windows are the time of the day and the ambient temperature