# AI challenge for safe and low carbon power grid operation

Source: DOAJ / Energy and AI metadata; ScienceDirect DOI landing page; ResearchGate public full-text search snippet checked for pre-proof availability.
Retrieved: 2026-06-14

## Bibliographic Metadata

- Authors: Adrien Pavão; Antoine Marot; Jules Sintes; Viktor Eriksson Möllerstedt; Laure Crochepierre; Karim Chaouache; Benjamin Donnot; Van Tuan Dang; Isabelle Guyon
- Venue/status: Energy and AI, volume 22, article 100564, 2025. DOI 10.1016/j.egyai.2025.100564.

## Source Links

- https://doi.org/10.1016/j.egyai.2025.100564
- https://doaj.org/article/9361a70944f94898b03a31bf6c4f5251
- https://www.sciencedirect.com/science/article/pii/S2666546825000965
- https://www.researchgate.net/publication/395030141_AI_challenge_for_safe_and_low_carbon_power_grid_operation

## Abstract / Retrieval Summary

Achieving carbon neutrality by 2050 will require power-grid operators to absorb unprecedented volumes of variable solar and wind generation while maintaining reliability. The paper reports on an L2RPN challenge based on 16 years of weekly scenarios, 832 scenarios total, on a 118-node grid under realistic constraints. It casts real-time grid operation as a Markov decision process; six teams combine heuristics, optimization, data scaling, supervised learning, and reinforcement learning. The paper analyzes the winning solution, reporting under-five-second multimodal actions, action-space reduction, a neural policy that predicts useful grid actions with over 80 percent accuracy, and a neural alert module trained on 315,000 samples with 93.9 percent recall for dangerous states.

## Retrieval Note

No arXiv version was found. ScienceDirect PDF access returned HTTP 403 in this environment; ResearchGate returned HTTP 1020 to curl. Stored as metadata-only rather than fabricating a PDF artifact.