LIVED - Long Device Level Energy Data

LIVED Data Set


LIVED stands for Long Device Level Energy Data and contains measurements collected from smart plugs multi-sensors as depicted. The data has been collected since 2011 and covers approximately 50 households. The typical deployment package of a household consists of seven smart plugs and four multisensors. The smart plugs are typically configured with a sampling rate of 2 seconds, thus delivering high resolution energy readings as compared to today’s smart meters that sample at intervals of 15 minutes.

Size of the Data Set

The data set encompasses more than 11 billion measurements.


We will realease a subset of the data [1] of 2.5 billion measurements covering 6 households over a period of two years in the next few weeks. in Hobbit we will use the full dataset and provide it as simulated data set as well.

Use Case Description

The data set has been used for various forms of analytics and data exploration including anomaly detection, activity recognition in various contexts. The primary scenarios for HOBBIT however is that of energy demand forecasting [2].

Structure of the data set

The readings are provided as tgz compressed csv files using ";" as a separator. the files are organized in subdirectories according to houseid, MAC address, sensor id, year and month. The folder names for MAC address and sensor id use URL encoding. For instance, in order to retrieve all the readings for the house with id 5, MAC address 00:00:00:00:00:00:00:19, sensor id 5:163 in May, 2015, you would access the files in the directory /houseid=5/mac=00%3A00%3A00%3A00%3A00%3A00%3A00%3A19/sensor_id=5%3A163/year=2013/month=5/. The directory structure and data format is the results of an Impala managend table.

The fields are as follows

  • ts (timestamp): a timestamp representing the time when the measurement was made
  • type: a string representing the measurement type (see Table 1).
  • value: a string representing the measurement value
  • unit: a string describing the unit of measurement
  • houseid: an integer representing the house
  • mac: MAC address of the device (smart plug or multisensor) containing the sensors. The MAC address is an artificially created MAC address and does not contain any information about the real MAC addresses used.
  • sensor_id: string identifying the sensor. The sensor identifier contains the houseid concatenated by a unique sensor identifier.
  • year: year identifying the year of the measurement (This is not actually stored, but part of the partitioning information for efficient access)
  • month: month identifying the month of the measurement (This is not actually stored, but part of the partitioning information for efficient access)


  1. Leonardi, A.; Ziekow, H.; Strohbach, M.; Kikiras, P. Dealing with Data Quality in Smart Home Environments—Lessons Learned from a Smart Grid Pilot. J. Sens. Actuator Netw. 2016, 5, 5. doi:10.3390/jsan5010005
  2. Ziekow, H.; Goebel, C.; Struker, J.; Jacobsen, H.A. The potential of smart home sensors in forecasting household electricity demand. In Proceedings of the 2013 IEEE International Conference on Smart Grid Communications (SmartGridComm), Vancouver, BC, Canada, 21–24 October 2013; pp. 229–234.

Data and Resources

Additional Info

Field Value
External Description
Contact LIVED Dataset Support
Contact Email
Benchmark Analysis and Processing