GAN-NILM tries to perform Non-Intrusive Load Monitoring using Generative Adversarial Networks. That is, in plain language, a model that takes only a building total measurement and tries to tell which appliances are On/Off at every time step.
The complete model and results is discussed in our paper: A. M. A. Ahmed, Y. Zhang and F. Eliassen, "Generative Adversarial Networks and Transfer Learning for Non-Intrusive Load Monitoring in Smart Grids," 2020 IEEE SmartGridComm
A summary can be found in this video presentaion: https://www.youtube.com/watch?v=Z_K5YEuSCOs&t=4s
The repository consists of ipynb files of the experiment already run. It still shows the outputs that were not deleted to give insigts about the results.
Datasets used can be found:
- REFIT: https://www.refitsmarthomes.org/datasets/
- UKDALE: https://jack-kelly.com/data/
- REDD: http://redd.csail.mit.edu/
Please cite:
A. M. A. Ahmed, Y. Zhang and F. Eliassen, "Generative Adversarial Networks and Transfer Learning for Non-Intrusive Load Monitoring in Smart Grids," 2020 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm), Tempe, AZ, USA, 2020, pp. 1-7, doi: 10.1109/SmartGridComm47815.2020.9302933.
BibTeX:
@INPROCEEDINGS{9302933,
author={A. M. A. {Ahmed} and Y. {Zhang} and F. {Eliassen}},
booktitle={2020 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGridComm)},
title={Generative Adversarial Networks and Transfer Learning for Non-Intrusive Load Monitoring in Smart Grids},
year={2020},
volume={},
number={},
pages={1-7},
doi={10.1109/SmartGridComm47815.2020.9302933}}