Residential Smart Meter Energy Time Series: Active power measurements with 1s reporting rate
The dataset includes active power measurements for a residential dwelling (apartment) located in Bucharest, Romania, collected at 1s second reporting rate over several months.Always-on appliances include the refrigerator and the wireless router. Several other appliances are installed in the residential unit: washing machine, lighting fixtures, electrical iron, vacuum cleaner, various ICT charging devices, and air conditioning (seldom used).
We hope that the dataset is useful to energy systems and computational intelligence researchers for time series forecasting, classification and energy disaggregation tasks.
For collecting the energy measurement data the Unbundled Smart Meter (USM) concept is used. The USM approach is a systematization where smart meter functionalities are adequately grouped into two separate (unbundled) components: (i) a module for metrological and hard real-time functions, called the Smart Metrology Meter (SMM), which has fixed (frozen) functionality and high security of recorded data (black box-like standard, where data can be lost only after buffer recirculation after known periods, e.g. 3 months or one year) and (ii) a Smart Meter eXtension (SMX) which has high flexibility to accommodate new functionalities, to be deployed during the meter lifetime and to support the future evolution of the smart grid and energy services.
The USM concept is described in detail in:
M. Sanduleac, L. Pons, G. Fiorentino, R. Pop and M. Albu, "The unbundled smart meter concept in a synchro-SCADA framework," 2016 IEEE International Instrumentation and Measurement Technology Conference Proceedings, 2016, pp. 1-5, doi: 10.1109/I2MTC.2016.7520459.
A data analytics approach using this data set for time series data mining using the Matrix Profile technique for feature extraction is presented in:
G. Stamatescu, R. Plamanescu, A. -M. Dumitrescu, I. Ciomei and M. Albu, "Multiscale Data Analytics for Residential Active Power Measurements through Time Series Data Mining," 2022 IEEE 7th International Energy Conference (ENERGYCON), 2022, pp. 1-5, doi: 10.1109/ENERGYCON53164.2022.9830170.
The zipped folder includes subfolders with text files with active power readings in Watts, corresponding to daily time series for several complete months from the year 2020: January, February, April, May, June, July, August, September. Each record (line) in the files contains a timestamp ("%m/%d/%Y %H:%M:%S %f" format) and the active power value.