An Analytical Federated Learning Method for Scalable Nonintrusive Load Monitoring
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 2513218-2513218
DOI: 10.1109/tim.2026.3704092 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Wenlong Guo und 5 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Nonintrusive load monitoring (NILM) infers appliance-level information from aggregate signals, which is expected to be achieved by federated learning over large-scale decentralized data. However, cross-user heterogeneity in appliance types and operating patterns induces strong nonindependent and identically distributed (non-IID) characteristics. Therefore, existing methods suffer from global–local bias, leading to performance degradation and high computational and communication costs as the number of users increases. To address these challenges, we propose an analytical federated learning method for scalable NILM. We first pretrain a shared feature extractor on simulated data to learn general waveform priors without precollected real-world data. The extractor is then frozen locally and analytically adapted through an update branch. We further derive a closed-form cloud aggregation that is equivalent to centralized learning in a single communication round, thereby avoiding global–local bias. Furthermore, our method is optimized for edge-side computation and communication to reduce resource consumption. Experiments across five datasets demonstrate superior performance, comparable to centralized learning even at 800 users. In addition, deployment on STM32 validates real-world efficiency, completing federated learning within 6.8 min. The code is publicly available at: https://github.com/2771618309/An-Analytical-Federated-Learning-Method-for-Scalable-NILM
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Guo, W., Yu, T., Luo, Q., Hu, X., Wu, Y., & Pan, Z. (2026). An Analytical Federated Learning Method for Scalable Nonintrusive Load Monitoring. IEEE Transactions on Instrumentation and Measurement, 75, 2513218-2513218. https://doi.org/10.1109/tim.2026.3704092
Vancouver
Guo W, Yu T, Luo Q, Hu X, Wu Y, Pan Z. An Analytical Federated Learning Method for Scalable Nonintrusive Load Monitoring. IEEE Transactions on Instrumentation and Measurement. 2026;75:2513218-2513218. doi:10.1109/tim.2026.3704092
BibTeX
@article{Guo2026,
author = {Guo, Wenlong and Yu, Tao and Luo, Qingquan and Hu, Xiaolei and Wu, Yufeng and Pan, Zhenning},
title = {An Analytical Federated Learning Method for Scalable Nonintrusive Load Monitoring},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {2513218--2513218},
doi = {10.1109/tim.2026.3704092}
}
EUVIMED