The open dataset described in this paper contains several years of low-resolution building energy consumption data collected from 17 buildings located in different European countries, and Australia. Collectively, the 17 buildings comprise data from 371 energy meters, which sampled data at frequencies within the interval of minutes to one hour, resulting in approximately 15 billion data points. Originally created for the 'ADRENALIN 2024: Building Energy Load Disaggregation Challenge', the dataset contains time-series main and sub-meter data, corresponding to the total energy consumption (main meters for the whole-building), and to the weather-dependent heating and cooling loads (sub-meters), respectively. The dataset also contains contextual data of the buildings, together with relevant weather data for the whole period. In addition to the variability in location, the dataset includes variability in the primary use of the buildings, including housing and non-residential. Besides load disaggregation, the dataset could also be used for other energy-related research purposes, including training of algorithms aimed to power flexibility services and even smart control of applications.
Bengtsson et al. (Mon,) studied this question.