Non-intrusive Appliance Load Monitoring (NALM) is essential for efficient electricity consumption tracking in households, promoting eco-friendly practices, and cost reduction. Privacy concerns in real-world NALM implementations can be addressed using federated learning. However, challenges persist, such as limited training data, edge model customization, and resource constraints. We introduce a practical federated learning framework for NALM, leveraging cloud-based model compression, collaborative data collection, and personalized edge and multi-task learning models through unsupervised transfer learning. Our advanced load monitoring model offers precise energy disaggregation through real-world experimentation, making it a leading solution for non-intrusive appliance load monitoring on edge devices. It maintains excellent learning performance and user privacy. Future research should aim to enhance federated learning efficiency and tackle remaining challenges in practical NALM implementations
%0 Conference Paper
%1 Natarajan2024
%A Natarajan, Yuvaraj
%A Wadhwa, Gitanjali
%A K R, Sri Preethaa
%A Sundaram, Karthic
%A K, Rama Abirami
%B 2024 International Conference on Green Energy, Computing and Sustainable Technology (GECOST)
%D 2024
%K imported yaff
%P 386-391
%R 10.1109/GECOST60902.2024.10474899
%T Advancing Sustainable IoT Appliance Load Monitoring Through Edge-Enabled Federated Transfer Learning
%X Non-intrusive Appliance Load Monitoring (NALM) is essential for efficient electricity consumption tracking in households, promoting eco-friendly practices, and cost reduction. Privacy concerns in real-world NALM implementations can be addressed using federated learning. However, challenges persist, such as limited training data, edge model customization, and resource constraints. We introduce a practical federated learning framework for NALM, leveraging cloud-based model compression, collaborative data collection, and personalized edge and multi-task learning models through unsupervised transfer learning. Our advanced load monitoring model offers precise energy disaggregation through real-world experimentation, making it a leading solution for non-intrusive appliance load monitoring on edge devices. It maintains excellent learning performance and user privacy. Future research should aim to enhance federated learning efficiency and tackle remaining challenges in practical NALM implementations
@inproceedings{Natarajan2024,
abstract = {Non-intrusive Appliance Load Monitoring (NALM) is essential for efficient electricity consumption tracking in households, promoting eco-friendly practices, and cost reduction. Privacy concerns in real-world NALM implementations can be addressed using federated learning. However, challenges persist, such as limited training data, edge model customization, and resource constraints. We introduce a practical federated learning framework for NALM, leveraging cloud-based model compression, collaborative data collection, and personalized edge and multi-task learning models through unsupervised transfer learning. Our advanced load monitoring model offers precise energy disaggregation through real-world experimentation, making it a leading solution for non-intrusive appliance load monitoring on edge devices. It maintains excellent learning performance and user privacy. Future research should aim to enhance federated learning efficiency and tackle remaining challenges in practical NALM implementations},
added-at = {2025-01-20T14:48:11.000+0100},
author = {Natarajan, Yuvaraj and Wadhwa, Gitanjali and K R, Sri Preethaa and Sundaram, Karthic and K, Rama Abirami},
biburl = {https://puma.scadsai.uni-leipzig.de/bibtex/22fc8d8e621c3daa7142fe5e4c1728f48/scadsfct},
booktitle = {2024 International Conference on Green Energy, Computing and Sustainable Technology (GECOST)},
doi = {10.1109/GECOST60902.2024.10474899},
interhash = {a41d9acc3715cd1d1f455de13b771576},
intrahash = {2fc8d8e621c3daa7142fe5e4c1728f48},
keywords = {imported yaff},
month = jan,
pages = {386-391},
timestamp = {2025-07-29T10:29:43.000+0200},
title = {Advancing Sustainable IoT Appliance Load Monitoring Through Edge-Enabled Federated Transfer Learning},
year = 2024
}
