EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Latent Skill Mining for Distribution Grid Switching Policy Recovery from Historical Operational Switching Logs

Elliot Hamilton, Maya Turner

Journal of Clinical Practice and Health Sciences · 10.06.2026 · 6 (1) · 173-183

DOI: 10.71465/jcphs57 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Clinical Practice and Health Sciences, online erschienen am 10.06.2026. Verfasst von Elliot Hamilton und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing integration of distributed energy resources and the electrification of various sectors have introduced unprecedented complexities into modern distribution grids. Managing these complexities often requires dynamic topology reconfiguration through switching operations, which are traditionally handled by human operators or rule-based systems. As grid states become more volatile, there is a critical need to automate these switching policies using advanced machine learning techniques. This paper introduces a comprehensive framework based on latent skill mining for the recovery and optimization of distribution grid switching policies from historical operational data. By treating grid reconfiguration as a sequential decision-making process, we propose a methodology that extracts reusable, high-level behavioral primitives, referred to as latent skills, from offline datasets. These skills represent fundamental topological transition strategies that human operators intuitively deploy to resolve congestion, isolate faults, and minimize power losses. The proposed approach leverages a hierarchical architecture where a high-level policy selects continuous latent skills, and a low-level policy translates these skills into specific sequences of switch actions. Through extensive analysis and textual description of the algorithmic structures, this paper demonstrates how latent skill mining mitigates the challenges of large discrete action spaces and distribution shifts commonly found in standard imitation learning and offline reinforcement learning. The findings suggest that the recovered policies exhibit superior adaptability and robustness in unobserved grid conditions compared to traditional monolithic models.

Zitieren

APA

Hamilton, E., & Turner, M. (2026). Latent Skill Mining for Distribution Grid Switching Policy Recovery from Historical Operational Switching Logs. Journal of Clinical Practice and Health Sciences, 6(1), 173-183. https://doi.org/10.71465/jcphs57

Vancouver

Hamilton E, Turner M. Latent Skill Mining for Distribution Grid Switching Policy Recovery from Historical Operational Switching Logs. Journal of Clinical Practice and Health Sciences. 2026;6(1):173-183. doi:10.71465/jcphs57

BibTeX
@article{Hamilton2026,
  author = {Hamilton, Elliot and Turner, Maya},
  title = {Latent Skill Mining for Distribution Grid Switching Policy Recovery from Historical Operational Switching Logs},
  journal = {Journal of Clinical Practice and Health Sciences},
  year = {2026},
  volume = {6},
  number = {1},
  pages = {173--183},
  doi = {10.71465/jcphs57}
}
Kostenlos starten →