RESIDENTIAL APPLIANCE USAGE PATTERNS FROM OVERALL ENERGY CONSUMPTION DATA: A STATISTICAL MACHINE LEARNING APPROACH
Conference Paper
Overview
Research
Identity
Other
View All
Overview
abstract
AbstractLarge energy-intensive domestic appliances like heating, ventilation, and air-conditioning systems, electric water heaters, refrigerators, clothes washers, dishwashers, and dryers can together comprise about 58% of annual residential electricity consumption in the US. Accurate knowledge of domestic appliance-level energy usage patterns can help energy modelers and electric utilities design optimal demand response programs in residential communities. However, while the widespread installation of residential smart meters in the US over the past decade has enabled electric utilities to collect diurnal individual household-level electricity consumption profiles, it is extremely rare to find location-specific, short-interval, disaggregated appliance-level data for individual residences. This study develops a novel methodology to reliably predict future domestic appliance energy consumption profiles based primarily on overall electricity consumption and local weather data. The model is demonstrated using fifteen-minute interval appliance-level empirical energy consumption data for a test-bed of 25 single-family detached homes in Austin, TX from Pecan Street Inc. a non-profit entity based in Austin. Although the training of our model utilizes historical appliance profiles, the results obtained from this analysis can be used to reliably and accurately predict time-granular appliance-level energy consumption patterns (not only on/off times) using only the overall electricity consumption profile of the household. Thus, this study opens up the possibility for energy modelers to reliably forecast domestic appliance electricity consumption profiles without having appliance-level historical energy usage datasets available.
