SMITE: using smart meters to infer the thermal efficiency of residential homes

Residential homes represent approximately 22 % of global energy use and a large proportion of this is due to space heating. The thermal efficiency of a building is typically evaluated manually via surveys, or via intrusive measurements requiring homes to be vacated for prolonged periods, which can result in great inconvenience and expense. More recently, non-intrusive methods have been developed to infer the thermal efficiency of a home which reduces the time and cost of identifying where interventions, such as installing insulation, will have the greatest impact in reducing heating energy usage and carbon emissions. The insight into thermal energy efficiency can also be used as a tool to help support those who are identified as fuel poor. However, none of the current non-intrusive methods take advantage of the half-hourly smart-meter readings that are presently available. This paper proposes a novel algorithm, SMITE, that detects the time periods of the day where the heating of a home is on for an extended length of time and uses this selected data to infer the heating loss coefficient (HLC) and the heating power loss coefficient (HPLC) of the home. The SMITE method is evaluated on 7 homes where the HLC has been inferred by a co-heating test and compared to a state-of-the-art non-intrusive algorithm for inferring HLC, Deconstruct. Our method shows a significant improvement when there is gas heating, with the mean absolute percentage error (MAPE) between the inferred and the co-heating HLC value reducing from 32.6 % for the Deconstruct method to 12.0% for the SMITE method. This paper also discusses the merits of using the HPLC (instead of the HLC) as an industry standard for evaluating thermal efficiency