EUVIMEDEUVIMED entdecken
Crossref

Large Language Model Guided Geographic Data Enrichment across Infrastructure Domains for Resilient Energy System Design

Jakob Geiges, Gökhan Demirel, Julian Hoffmann, Thorsten Schlachter, Kevin Förderer, Veit Hagenmeyer

ACM SIGEnergy Energy Informatics Review · 01.09.2026 · 6 (3) · 269-281

DOI: 10.1145/3861130.3861151 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACM SIGEnergy Energy Informatics Review, online erschienen am 11.11.2026. Verfasst von Jakob Geiges und 5 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Precise geographical coordinates of energy infrastructure entities are essential for power-system modeling, grid planning, and resilience analysis, yet major registries remain spatially incomplete and conventional geocoders often fail on ambiguous or weakly documented facilities. We evaluate a structured Large Language Model (LLM)-based enrichment pipeline that combines parametric prompting, optional web retrieval, and repeated-call medoid aggregation across four domains: European power plants, European airports, US hospitals, and global wastewater treatment plants. Web-augmented configurations substantially outperform the conventional OpenStreetMap-based Nominatim baseline on power plant entities, with GPT5 nano (web-augmented) localizing 87.5% of sampled power plants within 1 km at approximately $0.0038 per entity, and the same configuration generalizes strongly across the other domains. Parametric knowledge alone already recovers highly visible entities such as airports, while open-weight models improve recall over Nominatim but remain weaker at sub-kilometer precision. Overall, LLMs are promising candidate generators for registry completion, but raw confidence scores are not calibrated probabilities and high-accuracy applications still require retrieval support plus downstream validation.

Zitieren

APA

Geiges, J., Demirel, G., Hoffmann, J., Schlachter, T., Förderer, K., & Hagenmeyer, V. (2026). Large Language Model Guided Geographic Data Enrichment across Infrastructure Domains for Resilient Energy System Design. ACM SIGEnergy Energy Informatics Review, 6(3), 269-281. https://doi.org/10.1145/3861130.3861151

Vancouver

Geiges J, Demirel G, Hoffmann J, Schlachter T, Förderer K, Hagenmeyer V. Large Language Model Guided Geographic Data Enrichment across Infrastructure Domains for Resilient Energy System Design. ACM SIGEnergy Energy Informatics Review. 2026;6(3):269-281. doi:10.1145/3861130.3861151

BibTeX
@article{Geiges2026,
  author = {Geiges, Jakob and Demirel, Gökhan and Hoffmann, Julian and Schlachter, Thorsten and Förderer, Kevin and Hagenmeyer, Veit},
  title = {Large Language Model Guided Geographic Data Enrichment across Infrastructure Domains for Resilient Energy System Design},
  journal = {ACM SIGEnergy Energy Informatics Review},
  year = {2026},
  volume = {6},
  number = {3},
  pages = {269--281},
  doi = {10.1145/3861130.3861151}
}
Kostenlos starten →