Physics-Informed Machine Learning for Sustainable Power System Resilience Under Extreme Events: Deployable Intelligence for Low-Carbon Energy Infrastructure
IEEE Energy Sustainability Magazine · 2026-11 · 2 (4) · 19-32
DOI: 10.1109/esm.2026.3717560 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Energy Sustainability Magazine, 2026-11. Verfasst von Bingzhao Qiu und 5 weiteren Autor:innen. Die Arbeit nennt 6 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Extreme weather is redefining the operating limits of modern power systems and posing growing challenges to the low-carbon energy transition. For grids with high shares of renewable energy, resilience is inherently linked to sustainability, requiring power systems to maintain critical services, integrate distributed energy resources, avoid prolonged dependence on backup supply with high carbon intensity, and recover efficiently under limited emergency resources. Although machine learning has shown promise in weather impact forecasting, outage prediction, resource dispatch, and restoration planning, purely data-driven models remain difficult to deploy in mission-critical grid operations because extreme event data are scarce, tail risk extrapolation is unreliable, and unconstrained control policies may violate physical operating limits. These limitations are especially evident in ice storms, where weather, ice accretion, grid operation, deicing actions, and restoration processes are tightly coupled. This article proposes a physics-informed machine learning (PIML) framework for resilience and sustainability, termed PIML-R, to integrate data-driven intelligence, physically constrained grid operation, and sustainability-oriented resilience objectives. The framework consists of three deployable engines: GridAtlas for modular grid, hazard, resource, and sustainability indicator representation; GridOracle for digital twin simulation of multiple physical processes; and GridOptimizer for safety-aware, resource-efficient, and low-carbon decision optimization. Using preventive deicing under ice storm conditions as a representative case, the article demonstrates how PIML-R supports coordinated decision making before, during, and after extreme events. Finally, key challenges and future directions are outlined to advance autonomous, trustworthy, socially responsible, climate-adaptive, sustainable, and resilient grids.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Qiu, B., Yang, P., Ma, L., Yao, T., Liu, Y., & Ban, M. (2026). Physics-Informed Machine Learning for Sustainable Power System Resilience Under Extreme Events: Deployable Intelligence for Low-Carbon Energy Infrastructure. IEEE Energy Sustainability Magazine, 2(4), 19-32. https://doi.org/10.1109/esm.2026.3717560
Vancouver
Qiu B, Yang P, Ma L, Yao T, Liu Y, Ban M. Physics-Informed Machine Learning for Sustainable Power System Resilience Under Extreme Events: Deployable Intelligence for Low-Carbon Energy Infrastructure. IEEE Energy Sustainability Magazine. 2026;2(4):19-32. doi:10.1109/esm.2026.3717560
BibTeX
@article{Qiu2026,
author = {Qiu, Bingzhao and Yang, Peipei and Ma, Liyang and Yao, Tingting and Liu, Yiqi and Ban, Mingfei},
title = {Physics-Informed Machine Learning for Sustainable Power System Resilience Under Extreme Events: Deployable Intelligence for Low-Carbon Energy Infrastructure},
journal = {IEEE Energy Sustainability Magazine},
year = {2026},
volume = {2},
number = {4},
pages = {19--32},
doi = {10.1109/esm.2026.3717560}
}
EUVIMED