Information The Lerta Electrical Load Measurements dataset is dedicated for advancing the research of non-intrusive load monitoring methods. It includes electrical consumption data at aggregated and appliance level. It was collected in 4 households in Poland (Estern Europe) and covers a period of one and a half years. The data is provided in a comma-separated values format in a similar way to REFIT dataset. Both raw and cleaned data (6 seconds sample interval) is provided alongside code that shows raw to cleaned conversion and allows loading it into a nilmtk library. Acknowledgments The data was gathered during a project “Lerta – system inteligentnego opomiarowania zużycia oraz bezinwazyjnej analizy poboru energii elektrycznej i gazu w gospodarstwach domowych” (project POIR.01.01.01-00-0700/17) funded by The National Centre for Research and Development (call for proposal number: 3/1.1.1/2017). Licensing This work is licensed under the Attribution-ShareAlike 4.0 International (CC BY-SA 4.0). Citation (COMING SOON) Please cite the following paper if you use the dataset: The paper describing the dataset alongside results obtained using it is being reviewed and the details will be provided below. COMING SOON - bibtech citation information For now, please attribute the company: LERTA SA (https://lerta.energy/) Details Data format (cleaned) A csv file with following columns: time, aggregated, [appliance_name / appliance_id], …, [appliance_name / appliance_id] Each row contains following values: timestamp and energy usage in Watts. Data cleaning Details of light data cleaning procedure can be explored in a repository: https://github.com/husarlabs/nilm-dataset-converter In general the data was resampled to 6 seconds intervals. Data gathering procedure Data was gathered using Develco Smart Plugs (individual appliances) and Develco External Meter Interfaces (aggregated power demand) that send data to the cloud using Develco Gateway devices. Raw data was compressed in a way that only a usage above a threshold was registered. That means the raw data was not sampled in a regular way. nilmtk compatibility The dataset can be easily loaded into a popular nilmtk library. A converter was created and proposed to the nilmtk module. The code is available: https://github.com/husarlabs/nilmtk/tree/feature/lerta-converter