A Data-Driven Method for Unsupervised Electricity Consumption Characterisation at the District Level and Beyond

Abstract

A bottom-up electricity characterisation methodology of the building stock at the local level is presented. It is based on the statistical learning analysis of aggregated energy consumption data, weather data, cadastre, and socioeconomic information. To demonstrate the validity of this methodology, the characterisation of the electricity consumption of the whole province of Lleida, located in northeast Spain, is implemented and tested. The geographical aggregation level considered is the postal code since it is the highest data resolution available through the open data sources used in the research work. The development and the experimental tests are supported by a web application environment formed by interactive user interfaces specifically developed for this purpose. The paper's novelty relies on the application of statistical data methods able to infer the main energy performance characteristics of a large number of urban districts without prior knowledge of their building characteristics and with the use of solely measured data coming from smart meters, cadastre databases and weather forecasting services. A data-driven technique disaggregates electricity consumption in multiple uses (space heating, cooling, holidays and baseload). In addition, multiple Key Performance Indicators (KPIs) are derived from this disaggregated energy uses to obtain the energy characterisation of the buildings within a specific area. The potential reuse of this methodology allows for a better understanding of the drivers of electricity use, with multiple applications for the public and private sector.