Authors: Ellen Talbot*, University of Liverpool
Topics: Temporal GIS, Quantitative Methods
Keywords: Machine Learning, Energy, Smart Meters,
Session Type: Paper
Scheduler ID: THU-064-3:20 p.m.
Start / End Time: 3:20 PM / 5:00 PM
Room: Bayside A, Sheraton, 4th Floor
Domestic energy consumption continues to grow, as demand for powered devices that meet the conventions of comfort, convenience and cleanliness continues to rise (Hargreaves, Nye, & Burgess, 2010). In developed economies, smart metering is considered instrumental on many Government’s agenda for increased energy efficiency and more efficiently producing and storing resources to meet the demand through measurement and monitoring of consumer energy behaviour (Guerreiro, Batel, Lima, & Moreira, 2015). This research evaluates the potential for spatio-temporal signatures of aggregate residential behaviour to be extracted from domestic smart meter data in conjunction with socio-demographic determiners of energy consumption. It is hypothesised that temporal trends in energy consumption could be of utility when defining patterns of daily life and providing a real-time indicator of sources and syncs of residential populations at a highly granular level. A machine learning methodology is applied to the smart meter dataset to understand how well socio-demographic indicators represent energy usage across England and Wales at multiple temporalities. Consideration is given firstly to overall penetration rates of smart meters and it is anticipated that these results will help to explain the bias’s in consumer led big datasets, and give an insight into the spatial distribution of Smart Meters. This basic model is then reworked to understand how energy usage and the relevance of socio-demographic indicators are affected at the seasonal, diurnal and hourly granularities.