HG-RAG: Hierarchical Graph-Enhanced Retrieval-Augmented Generation for Power Systems
Electronics · 30.03.2026 · 15 (7) · 1445
DOI: 10.3390/electronics15071445 ↗Kurz gefasst
Fachartikel in der Zeitschrift Electronics, online erschienen am 30.03.2026. Verfasst von Zhijun Shen und 5 weiteren Autor:innen. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Science and Technology Program of China Southern Power Grid Co., Ltd., Power Dispatching and Control Center of Guangdong Power Grid Co., Ltd.
Abstract (Originalsprache)
Retrieval-augmented generation (RAG) has shown strong potential for knowledge-intensive tasks, yet its performance degrades sharply when applied to structured long-context documents in power systems, where dense entity–relation dependencies, cross-document references, and strict traceability requirements exist. To address this Structured Long-Context RAG (SLCRAG) challenge, this paper proposes a hierarchical graph-enhanced RAG (HG-RAG) framework tailored for power system question answering. HG-RAG constructs a globally consistent knowledge graph via sliding-window entity–relation extraction to mitigate semantic fragmentation, and employs multi-granularity structured indexing for precise entity/relation retrieval. A hierarchical structured retrieval mechanism with multi-hop expansion and semantic distillation maximizes recall while minimizing redundancy. Furthermore, a regex-enhanced retrieval module records authoritative file_path provenance and constrains downstream retrieval to the same source documents, effectively eliminating cross-document interference—especially in cases where different documents contain similar entities and relations. Combined with version control and deduplication-merging, HG-RAG supports incremental knowledge updates with minimal forgetting and negligible token overhead. Experimental results on a domain-authentic power system QA dataset demonstrate that HG-RAG outperforms LightRAG and GraphRAG, achieving up to 85.47% accuracy in short-answer tasks with significantly lower token consumption. Ablation studies confirm that semantic distillation primarily improves precision and efficiency, while regex-enhanced retrieval safeguards recall in edge cases.
Förderung
- the Science and Technology Program of China Southern Power Grid Co., Ltd.
- Power Dispatching and Control Center of Guangdong Power Grid Co., Ltd
Zitieren
APA
Shen, Z., Cai, X., Ni, B., Meng, Z., Huang, Z., & Yu, T. (2026). HG-RAG: Hierarchical Graph-Enhanced Retrieval-Augmented Generation for Power Systems. Electronics, 15(7), 1445. https://doi.org/10.3390/electronics15071445
Vancouver
Shen Z, Cai X, Ni B, Meng Z, Huang Z, Yu T. HG-RAG: Hierarchical Graph-Enhanced Retrieval-Augmented Generation for Power Systems. Electronics. 2026;15(7):1445. doi:10.3390/electronics15071445
BibTeX
@article{Shen2026,
author = {Shen, Zhijun and Cai, Xinlei and Ni, Binye and Meng, Zijie and Huang, Zhanhong and Yu, Tao},
title = {HG-RAG: Hierarchical Graph-Enhanced Retrieval-Augmented Generation for Power Systems},
journal = {Electronics},
year = {2026},
volume = {15},
number = {7},
pages = {1445},
doi = {10.3390/electronics15071445}
}
EUVIMED