FIRED

Beteiligte Mitarbeiter
Lehrstuhl für Rechnerarchitektur
Benjamin Voelker, M.Sc.	Entwickler

Projektbeschreibung

We present the FIRED dataset. It contains 101 days of 8kHz aggregated current and voltage readings of the 3-phase supply of a typical residential apartment in Germany. The dataset also contains matching ground truth data as 2kHz readings of 21 individual appliances. Furthermore, it includes sensor readings such as room temperatures and fully labeled state changes of the lighting resulting in the most complete and versatile residential electricity dataset currently known to us.

Software

We provide python tools that allow to load the dataset and extract some statistics as open source.
Link: https://github.com/voelkerb/FIRED_dataset_helper

Dataset

The dataset contains roughly about 2.4TB of data. In addition to the raw 8 and 2kHz current and voltage waveform data (2.4TB), we provide both a 1Hz and a 50Hz summary of the data. The summary contains active, reactive and apparent power derived from the raw data. The 50Hz summary is roughly about 30GB while the 1Hz data only has around 2GB and allows to get quick insight into the data.

Data records

High Freq Data

High frequency current and voltage waveforms are available for the appartments mains and 21 individual appliances. The mains are recorded with 8kHz. Individual appliances have been recorded with 2kHz.

Summary Data:

A dataset summary is available. Active, reactive and apparent power has been derived from the raw current and voltage waveforms and stored in a 50Hz and downsampled 1Hz version. The 1Hz version allows to get quick insights into the data.

File Format:

All data is stored inside matroska multimedia container. Data is compressed using 32 bit lossless wavpack compression.

Each file has the form:

"recordingDeviceIdentifier_year_mon_date__hour_min_sec.mkv"

Units are Volt, milli Ampere and Watt (W, var, VA) for voltage, current and power (active, reactive and apparent) readings respectively.

Annotation Data:

State changes for the apartments smart lighting, television and hifi system is available in CSV format. The first column represent the timestamp associated with the event. The remaining colum(s) represent data of the event such as lighting dimm state or color.

Addiitional temperature and humidity readings are available for three rooms of the apartment.

Appliance Info:

Information about each electrical appliance is available as a machine readable JSON file and a human readable PDF file.

Download

See https://github.com/voelkerb/FIRED_dataset_helper

Publikationen

B. Völker, M. Pfeifer, P. M. Scholl and B. Becker. "FIRED: A Fully-labeled hIgh-fRequency Electricity Disaggregation Dataset."TBP: International Conference on Systems for Energy-Efficient Built Environments 2020.