Published September 10, 2024 | Version v1
Computational notebook Open

DSUALM10H

  • 1. Universidad de Almeria

Description

In this repository, the analysis of the measurements of 10 applications including harmonics up to order 150 of voltage, current and power is presented in the attached notebooks. The measurements are carried out with 3 three-phase OpenZmeters (each with 4 measurement channels), making a total of 11 measurement channels distributed among the 10 applications, plus the aggregate that compiles the accumulated measurements of all of them.

The measurements correspond to W, VAR, VA, f, VLN, PF and A, plus the harmonics up to order 50 of W, V and A, all with a 13-digit UNIX Epox timestamp. The generation of the Dataset was done with the new converter/converter designed for this occasion, starting from 11 csv files of measurements that resulted from joining each csv file of each application of each day with the csv of that same application and that same day.

DUE TO ITS LARGE SIZE, BECAUSE OF ITS EXTENSION, THE DATA FILES ARE NOT AVAILABLE IN CSV FORMAT IN THIS REPOSITORY, BUT IT IS AVAILABLE IN THIS REPOSITORY DSUALM10H THE COMPLETE DATASET IN COMPRESSED FORM IN RAR FORMAT WITH ALL THE HARMONICS WITH THE USUAL TIME OF SAMPLES.

These data were trained with both the combinatorial algorithm (CO) and the Hidden Markov algorithm (FHMM), but the algorithm that returned the best results was CO, since with FHMM it was impossible to run it with sampling times of less than 90seconds due to lack of physical memory (it even returned errors using machines with more than 64GB of RAM).

Our goal is to provide NILM researchers with new data repositories to expand the existing range. As these new datasets may contain more than 150 electrical variables recorded at high frequency in different everyday applications, by offering this wide range of data, we hope to boost and improve research in the NILM field.

The main conclusions of taking only the odd harmonics, disregarding all harmonics, taking all harmonics (even and odd), or extending the sample time are presented in generic form below.

Files

DSUALM10H_OZM-main.zip

Files (96.2 MB)

Name Size Download all
md5:21998f90dea8102c7065725b2e7e2956
96.2 MB Preview Download

Additional details

Software

Repository URL
https://github.com/crn565/DSUALM10H_OZM
Programming language
Python