Enabling Sustainable Smart Grids With Programmable Networking and Federated Learning: A Phasor Measurement Unit Missing-Data Recovery Case Study
IEEE Energy Sustainability Magazine · 2026-11 · 2 (4) · 48-58
DOI: 10.1109/esm.2026.3717261 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Energy Sustainability Magazine, 2026-11. Verfasst von Xiaoyu Guo und 1 weiteren Autor:innen. Die Arbeit nennt 5 Literaturangaben. Gefördert von: National Science Foundation, U.S. Department of Energy.
Abstract (Originalsprache)
As electric grids integrate more renewable generation, distributed resources, and data-driven control, sustainability increasingly depends on reliable communication and high-quality measurement data. Modern smart grids rely on continuous streams from devices, such as phasor measurement units (PMUs), to support real-time monitoring, renewable integration, energy efficiency, and grid stability. However, communication failures and cyberattacks can delay, corrupt, or interrupt critical measurements, degrading situational awareness and forcing more conservative operation. In renewable-rich grids, this loss of visibility can reduce the effective use of flexible and low-carbon resources. This article explores a sustainable and resilient smart grid intelligence architecture that combines programmable networking with federated learning. The Programming Protocol-Independent Packet Processors (P4) language enables the network to observe and react to selected traffic behaviors in real time, while federated learning improves models across distributed sites without transferring all measurement data to a central location. Together, they provide responsiveness, adaptability, and data locality for future smart grid communication systems. We use PMU missing-data recovery as a representative case study. A central controller coordinates model distribution and aggregation across 12 PMU measurement sites, while each site fine-tunes the shared model using local PMU data. This design preserves data locality, enables early awareness of communication-side disruptions, and supports collaborative model improvement. Experimental results show that the federated learning neural network (FLNN) achieves lower recovery errors than comparison methods, particularly for longer continuous gaps. More reliable PMU continuity can support state estimation, event detection, earlier corrective action, and more confident integration of variable renewable generation. Overall, this article shows how programmable data planes and distributed learning can support measurement continuity, communication resilience, and sustainable smart grid operation.
Förderung
- National Science Foundation
- U.S. Department of Energy
Zitieren
APA
Guo, X., & Jin, D. (2026). Enabling Sustainable Smart Grids With Programmable Networking and Federated Learning: A Phasor Measurement Unit Missing-Data Recovery Case Study. IEEE Energy Sustainability Magazine, 2(4), 48-58. https://doi.org/10.1109/esm.2026.3717261
Vancouver
Guo X, Jin D. Enabling Sustainable Smart Grids With Programmable Networking and Federated Learning: A Phasor Measurement Unit Missing-Data Recovery Case Study. IEEE Energy Sustainability Magazine. 2026;2(4):48-58. doi:10.1109/esm.2026.3717261
BibTeX
@article{Guo2026,
author = {Guo, Xiaoyu and Jin, Dong},
title = {Enabling Sustainable Smart Grids With Programmable Networking and Federated Learning: A Phasor Measurement Unit Missing-Data Recovery Case Study},
journal = {IEEE Energy Sustainability Magazine},
year = {2026},
volume = {2},
number = {4},
pages = {48--58},
doi = {10.1109/esm.2026.3717261}
}
EUVIMED