A large scale artificial dataset to evaluate the impact of on-board monitoring set-ups on the Heat Transfer Coefficient assessment

At the current state, the majority of residential buildings are underperforming in terms of energy. To assess the current state of the building sector and bridge the gap between the designed and operative energy demand, quantifying the as-built envelope behaviour is essential. For this purpose the Heat Transfer Coefficient is used as a stationary performance indicator, however, prevailing assessment methods are impractical as they are performed on individual buildings. To extend the assessment range to urban scale statistics-based modelling techniques can be applied to limited and on-board collected data. In this step, where the accuracy of statistical methods and different input scenarios are investigated, simulated data is more suitable since there is no influence of measurement inaccuracy and the expected outcome is known. As part of this research, a comprehensive dataset to assess the HTC estimation methodology is created. The artificial dataset is generated using urban building energy models representative of the Belgian residential building stock. The generated artificial datasets resemble ideal measurement campaigns where inputs can be systematically selected which supports the aim of determining the impact of available monitoring setups. In this work, black-box AutoRegressive with eXogenous input statistical models were applied and the results show that, considering the assumed modelling limitations, with ideal monitoring setups 85% of the estimates fall within 10% of the target.

Automated summary

Currently, most residential buildings are not performing well in terms of energy efficiency. Quantifying the behavior of the building envelope is crucial in assessing the current state of the building sector and bridging the gap between design and actual energy demand. To achieve this on an urban scale, statistics-based modeling techniques can be employed using limited and collected data. In this research, an artificial dataset is created to assess the estimation methodology of the Heat Transfer Coefficient (HTC) using urban building energy models. The results show that with optimal monitoring setups, 85% of the estimates fall within 10% of the target.