Predicting the diversity of internal temperatures from English residential sector using panel methods

In this paper, panel methods are applied in new and innovative ways to predict daily mean internal temperature demand across a heterogeneous domestic building stock over time. This research not only exploits a rich new dataset but presents new methodological insights and offers important linkages for connecting bottom-up building stock models to human behaviour. It represents the first time a panel model has been used to estimate the dynamics of internal temperature demand from the natural daily fluctuations of external temperature combined with important behavioural, socio-demographic and building efficiency variables. The model is able to predict internal temperatures across a heterogeneous building stock to within 0.71 C at 95% confidence and explain 45% of the variance of internal temperature between dwellings. The model confirms hypothesis from sociology and psychology that habitual behaviours are important drivers of home energy consumption. In addition, the model offers the possibility to quantify take-back (direct rebound effect) owing to increased internal temperatures from the installation of energy efficiency measures. The presence of thermostats or thermostatic radiator valves (TRVs) are shown to reduce average internal temperatures, however, the use of an automatic timer is shown to be statistically insignificant. The number of occupants, household income and occupant age are all important factors that explain a quantifiable increase in internal temperature demand. Households with children or retired occupants are shown to have higher average internal temperatures than households who do not. As expected, building typology, building age, roof insulation thickness, wall U-value and the proportion of double glazing all have positive and statistically significant effects on daily mean internal temperature. In summary, the model can be either used to make statistical inferences about the importance of different factors for explaining internal temperatures or as a predictive tool. However, a key contribution of this research is the possibility to use this model to calibrate existing building stock behaviour and socio-demographic effects leading to improved estimations of domestic energy demand