Smart electricity meters allow for capturing consumption data of individual households at a high resolution in time (typically at 15-minute intervals). The key objective of this project is to develop further and evaluate feature extraction and machine learning techniques for automatic identification of household properties based on electricity load profiles and additional consumption-related information (weather, socio-demographic data, holidays, etc.). The gained information shall render highly targeted and scalable energy efficiency services possible.
The developed classification methods enable recognition of 38 household characteristics with accuracy of partially above 70%, based on smart meter load profiles and additional freely available data and under adherence to data privacy and security regulations. The characteristics describe inhabitants’ life situation (e.g., families, retirees, children, social status), energy efficiency (e.g., heating type, age and size of house, appliances in the household) as well as attitudes (e.g., toward renewable energy sources, interest on green electricity or solar panels). The project results will help authorized energy service providers in realization of effective and scalable energy efficiency campaigns. At the same time, the results support a fact-based discussion of advantages (e.g., enhancement of energy efficiency) and costs (e.g., privacy implications) of such approaches
