08-Dec-2020: added new files to the Natural subset and updated corresponding sections in pdf User Guide.
About the Project
The LIT-Dataset was conceived and engineered to provide data for evaluation of NILM Systems.
One such Non-Intrusive Load Monitoring System (NILM) is under development in our Laboratory for Innovation and Technology in Embedded Systems (LIT).
The development process behind the LIT-Dataset started by the evaluation of exiting datasets, by comparing their features and then by stating the requirements of the recorded waveforms of voltage and current when single and multiple loads were monitored under several conditions.
A distinguished feature of the LIT-dataset is that it is composed of three distinct classes of load monitoring:
Synthetic: waveforms were collected with a jig that precisely controls the on and off times in a scenario of controlled load shaping. The jig controls each individual load, up to eight, using TRIACs and relays, hence, allowing for on-times at specified angles of the mains sine wave. The on and off times are recorded in the waveform with a resolution better that 5 ms, allowing the identification of the mains semi-cycle where the "on" or "off" load-event occurred. Typically waveforms are 30 seconds long at sampling rates above 15 KHz, achieving 256 samples per cycle.
Simulated: several classes of loads were simulated in a MATLAB/SIMULINK environment, and validated against actual loads. The use of simulated loads allows us to modify the load characteristics so to create scenarios that would be very difficult/complex to create in the real world.
Natural: waveforms were collected in an actual environment (residential, research, commercial, industrial) over longer periods of time. Sensors are attached to each load so to precisely detect and record the on and off times of each load, while the waveforms record the aggregate current and voltage of all monitored loads.
The LIT-Dataset is still under construction. New waveforms are added daily. All waveforms have the same password; registration is only needed once.
To reference our work, please use one or more of the following:
Each waveform is recorded in a set of files, some in binary format to reduce size, other in ASCII.
The format is specified in this arquive Specification file.pdf.
We also provide scripts so that these waveforms can be easily read into MATLAB and from there saved in many other formats.
For the reduced size files, we follow some patterns on naming the acquisitions: first, the number, which represents the amount of loads of an acquisition.
The letters that follow this number are the descriptions of which load we used in this acquisition.
See the tab Matlab scripts and guides.
The data here provided is being made available to the public interested in research in NILM as well as other topics.
All we ask is to reference our work. There are over a thousand waveform files, each one in a password protected zip file.
The next pages provide access to these files, organized by classes, type and number of concurrent loads.
We will mail the password to you once the registration form is filled out, all we ask is your name, affiliation, research topic and mail address. Register here.
This section presents a short preview of the sort of information to be obtained from the LIT-Dataset.
One of the downloadable files from the dataset is: sy1zip. From the list of loads, we can identify load N0 as being a 100W incandescent light. When expanding this zip file the files for load N0 will be in the folder: Synthetic/1/1N0.
After downloading and expanding the .zip file, the following files are available:
To facilitate the interpretation of these binary files, we provide a MATLAB script that converts these into .mat files with the voltage and current samples.
The 16 files shown correspond to 16 acquisitions of the same load (incandescent light) with varying power on angles along the mains cycle.
By plotting voltage (blue) and current (red) samples of one of these .mat files, we can see (left) the whole waveform and on the right a zoom into the power on.
Context and Funding
ANEEL is the Brazilian Electricity Regulatory Agency, an autarchy linked to the Ministry of Mines and Energy.
Copel is the power company operating in the state of Paraná, south of Brazil. Copel is responsible for Generation, Transmission, Distribution, Energy Trading and Telecommunications.
ANEEL is responsible for a national R&D funding program in the area of Electrical Energy. The execution of this program is performed by the utility companies. In 2017, UTFPR submitted to the Call for Proposals on the subject of “Energy Efficiency for Public Universities”. Project “A Methodology for Analysis, Monitoring and Management of Distributed Generation using Renewable Energy Sources” received a grant and is under development from 2017 to 2020. This project aims at Applied Research in the following topics:
On the Consumer Side:
Use of Power Signature technology to identify the individual consumption of electric loads.
On the Generation Side:
Monitoring and management of medium-scale grid-connected PV plants.
Solar radiation mapping and performance comparison of photovoltaic technologies in the state of Paraná-Brazil.
Development of a set of techniques to survey wind potential maps for metropolitan regions.
Development of a power flow analysis and a protection coordination model applicable to distributed generation in the rural electrification scenario.
The LIT Dataset is one of the results of this project.
Execution
Sponsors
This research is being supported by Copel Distribuição S.A., a Brazilian power company serving the state of Parana. Project PD 2866-0464-2017.
