Datasets
Open Access
Industrial Machines Dataset for Electrical Load Disaggregation
- Citation Author(s):
- Submitted by:
- Pedro Bandeira ...
- Last updated:
- Wed, 02/05/2020 - 18:03
- DOI:
- 10.21227/cg5v-dk02
- Data Format:
- License:
- Creative Commons Attribution
- Categories:
- Keywords:
Abstract
This dataset contains heavy-machinery data from the Brazilian industrial sector. The data was collected in a poultry feed factory located in the state of Minas Gerais, Brazil. Its process can be summarized to creating pellets of ration for poultry from corn or soybeans and added nutrients. The factory produces at fullscale over the entire year, thus it has well-behaved usage patterns at any time. It operates from Mondays through Fridays (and occasionally on Saturdays, in case production is below the monthly target) on a daily three-turn shift from 10:00 PM to 05:00 PM. From 05:00 PM to 10:00 PM electricity prices are higher, so the factory is closed.
The site meter collects data at the factory electrical substation from the main medium voltage/low voltage transformer (MV/LV), which transforms 13.4 kV to 380 V. The factory has four different LVDBs after the MV/LV. One (1) for lights and administration-related appliances, the second (2) for pelletizing-related machinery, the third (3) for milling-related machinery and the last one (4) for general production-related machinery. Eleven GreenAnt energy meters were installed on the factory: one as a site meter and the other ten for groundtruth analysis. These meters sample data at 8 kSamples/sec internally, downsample the data to 1 Hz. Each package includes: RMS voltage, RMS current, active power, reactive power and apparent power.
Samples were collected from 2017-12-11 18:43:52 UTC until 2018-04-01 21:33:17 UTC, which roughly corresponds to 111 days. Only one phase out of three was collected. We decided to do this as the factory is well balanced and would decrease the total amount of data stored by one third.
There are three meters measuring distribution circuits: MV/LV transformer, LVDB-2 and LVDB-3. This means that, there are eight GreenAnt meters on the factory measuring appliances, which are: Pelletizer I (PI); Pelletizer II (PII); Double-pole Contactor I (DPCI); Double-pole Contactor II (DPCII); Exhaust Fan I (EFI); Exhaust Fan II (EFII); Milling Machine I (MI); and Milling Machine II (MII). All machines under LVDB-2 and LVDB-3 work at 380 V and 60 Hz. The factory As the factory is cyclically producing pellets at full power, each appliance can be modeled as a threestate machine (OFF, NO LOAD ON, FULL LOAD ON). The milling machines were the last machines to be measured. They were only measured over the last 12 days, which corresponds to roughly 10% of the time measurement intervals from other machines.
Features
- RMS Voltage
- RMS Current
- Active Power
- Reactive Power
- Apparent Power
Appliances measured
- Pelletizer I
- Pelletizer II
- Double-pole Contactor I
- Double-pole Contactor II
- Exhaust Fan I
- Exhaust Fan II
- Milling Machine I
- Milling Machine II
Circuits Measured:
- Main Medium Voltage/Low Voltage transformer
- Pelletizers Sub-circuit (Low Voltage Distribution Board-2)
- Milling Machines Sub-circuit (Low Voltage Distribution Board-3)
Based on NILM METADATA.
Each CSV file is named according to the appliance or circuit it represents.
IMDELD.hdf5 is the complete dataset that uses NILM METADATA and is fully compatible with NILMTK.
Dataset Files
- Data from the MV/LV Transformer. mvlv-transformer.csv (247.02 MB)
- Pelletizer SubCircuit. pelletizer-subcircuit.csv (390.21 MB)
- Milling Machines SubCircuit. millingmachine-subcircuit.csv (75.16 MB)
- Zip file with all appliances. Appliances.zip (521.01 MB)
- Complete dataset compatible with NILMTK python module. IMDELD.hdf5 (2.70 GB)
Open Access dataset files are accessible to all logged in users. Don't have a login? Create a free IEEE account. IEEE Membership is not required.
Comments
Dear Author,
Thank you for sharing this dataset. Upon investigation the raw data (Data from the MV/LV Transformer) , I found several days of data is missing from the start to end date of measurement. Was there any measurement failure ? Is there any complete dataset from this study ?
Thank you again for you time and i look forward to hear from you
Kind regards
Shahnewaz