Impact of solar gain estimation on heat loss coefficient determination using in-situ data: Comparing co-heating test with B-splines integrated grey-box modelling

This study investigates the estimation uncertainties of grey-box models, with increased complexities of quantifying solar gain dynamics, in building energy performance assessment. The precision of solar gain estimation has been proven to be crucial for delivering a qualified grey-box model and achieving a better performance in predictive applications, e.g., model predictive control (MPC). Therefore, grey-box models that characterize solar gain dynamics more precisely are generally preferred. However, characterizing solar gain dynamics with more elaborated details might estimate solar gain dynamics more precisely but enlarge the uncertainties in accessing the buildings' overall energy performance, e.g., determining the buildings' overall heat loss coefficient (HLC). Hence, this study aims to understand this potential risk. At first, four types of solar gain estimation approaches available for grey-box models are systematically summarized. The four grey-box models are applied to two datasets monitored on a full-scale lightweight building, under two scenarios of effective window distribution. The study reveals that enhancing the solar gain modelling in grey box models does not hamper a reliable HLCdetermination in this case. Moreover, it was found that under certain conditions a more precise estimation of the solar gain can even reduce the uncertainty of the HLC-estimate via narrowing 95% confidence interval.

Automated summary

This study investigates the uncertainties in estimating solar gain dynamics in grey-box models for building energy performance assessment. The precision of solar gain estimation is crucial for a qualified grey-box model and improved predictive applications. The study found that enhancing solar gain modelling in grey-box models does not hinder reliable heat loss coefficient determination and can even reduce uncertainty in certain conditions.