Data-driven approach to prioritize residential buildings’ retrofits in cold climates using smart thermostat data

Abstract

At least 65% of existing residential buildings will still be in use by 2050, thus retrofitting existing buildings is critical to reducing energy consumption. However, prioritizing building retrofits typically requires a thorough evaluation of their thermal performance, which can be cost-prohibitive, especially on a large scale. To this end, this study presents a data-driven framework to target buildings for retrofits using smart thermostat data. To validate the framework, it was applied to 60,000 homes across North America using four years of real-time measurements. First, grey-box modelling approaches were used to estimate the thermal time constant for each home. Homes were then clustered according to their estimated values and for each cluster, the priority of retrofit was ranked. Finally, a classification model was developed to predict the priority of retrofit. Using a large sample size, the results can be used to prioritize buildings for retrofits when limited information is available. HIGHLIGHTS Thermostat data from over 60,000 houses were used to estimate their thermal performance. Two grey-box methods to estimate a building's thermal time constant (RC value) were compared. The estimated time constant values were used to cluster houses based on thermal performance. A classification model was developed to prioritize retrofits for each house based on its attributes.