Statistical models to infer gas end-use efficiency in individual dwellings using smart metered data

• A domestic gas end-use efficiency benchmarking tool is developed.
• A new method to estimate a heating degree day regression model is presented.
• The benchmarking tool is demonstrated using a sample of smart-metered consumers.

Residential buildings can significantly contribute to the European Union's 2020 efficiency energy targets. For this reason, energy distributors and suppliers are required to provide assistance to householders to reduce energy end-use. This paper develops statistical modelling methods that can be used by suppliers to infer the gas fuel efficiency of buildings in their residential portfolio, in order to deliver improved energy management services to consumers. The study begins by estimating individual statistical building energy models for a sample of consumers and presents the resulting distribution of independent parameters. These parameter distributions are then characterised by regression models using descriptive household data that is generally known by the consumer and can be easily gathered by the energy supply company. These models are then used to compare the inferred energy end-use efficiency of the household (cooking, hot-water and space heating) to similar dwellings. Buildings with higher-than-expected gas consumption can be targeted for energy efficiency programmes.