|本期目录/Table of Contents|

[1]徐浩,刘基灿,张焱,等.大模型赋能的光伏电站智能运维策略*——基于领域知识图谱的构建、补全及应用[J].中国安全生产科学技术,2026,22(3):197-204.[doi:10.11731/j.issn.1673-193x.2026.03.025]
　XU Hao,LIU Jican,ZHANG Yan,et al.Intelligent operation and maintenance strategies for photovoltaic power stations empowered by LLMs:construction,completion,and application of domain[J].Journal of Safety Science and Technology,2026,22(3):197-204.[doi:10.11731/j.issn.1673-193x.2026.03.025]
点击复制

大模型赋能的光伏电站智能运维策略*——基于领域知识图谱的构建、补全及应用

《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]

卷:
22
期数:
2026年3期
页码:
197-204
栏目:
公共安全与应急管理
出版日期:
2026-03-30

文章信息/Info

Title:
Intelligent operation and maintenance strategies for photovoltaic power stations empowered by LLMs:construction,completion,and application of domain
文章编号:
1673-193X(2026)-03-0197-08
作者:
徐浩; 刘基灿; 张焱; 邓三鸿; 康振渊
（1.南京工程学院 商学院，江苏 南京 211167;
2.南京大学 信息管理学院，江苏 南京 210023;
3.南京工程学院 电力工程学院-沈国荣学院，江苏 南京 211167）
Author(s):
XU Hao;  LIU Jican;  ZHANG Yan;  DENG Sanhong;  KANG Zhenyuan
(1.School of Business,Nanjing Institute of Technology,Nanjing Jiangsu 211167,China;
2.School of Information Management,Nanjing University,Nanjing Jiangsu 210023,China;
3.Nanjing Institute of Technology School of Electric Power Engineering,School of SHENGUORONG,Nanjing Jiangsu 211167,China)
关键词:
大语言模型; 光伏电站运维; 知识图谱补全; TuckER模型; 实体增强
Keywords:
large language models;  photovoltaic power plant operation and maintenance;  knowledge graph completion;  TuckER;  entity enhancement
分类号:
X913
DOI:
10.11731/j.issn.1673-193x.2026.03.025
文献标志码:
A
摘要:
为解决光伏电站运维领域知识碎片化、专业人才短缺以及故障诊断复杂等问题，提出大语言模型赋能的光伏电站运维知识图谱构建、补全与应用一体化策略，赋能光伏电站的智能运维。通过多轮Prompt迭代从光伏电站故障案例文本中自动抽取领域实体与关系，并构建领域知识图谱；以TuckER模型为基线，设计均值、加权与拼接3种邻居实体增强机制，融合实体自身与邻域信息以缓解图谱稀疏、表示质量低与补全性能弱等问题；从融合权重、训练数据规模与实体频率3个维度评估补全效果，进一步构建多跳推理决策框架。以“电池组件焊点熔化”为例验证诊断链的可行性与科学性。研究结果表明：相较于TuckER基线模型，本文提出的实体增强方法的MRR指标提升26.34%，对低频实体更友好的同时可有效缓解“冷启动”问题。研究结果可为故障知识自动获取、补全与可解释推理提供路径，有助于提升智能运维的效率。
Abstract:
In order to address fragmented knowledge,the shortage of skilled experts,and the complexity of fault diagnosis in photovoltaic power station operation and maintenance,this study proposes an integrated strategy,enabled by large language models,for domain knowledge graph construction,completion,and application,thereby supporting intelligent operation and maintenance in photovoltaic power stations.Through iterative prompt refinement,the proposed method automatically extracts domain entities and relations from fault case texts and constructs a domain knowledge graph.Using TuckER as the baseline model,it further develops three neighbor enhanced mechanisms,namely mean aggregation,weighted aggregation,and concatenation,which integrate an entity’s own representation with neighborhood information to alleviate graph sparsity,improve representation quality,and enhance completion performance.The effectiveness of knowledge graph completion is evaluated from three perspectives,including fusion weight,training data scale,and entity frequency,and a multi hop reasoning framework is further established for decision support.A case study on photovoltaic module solder joint melting verifies the feasibility and scientific soundness of the diagnostic reasoning chain.The results show that,compared with the TuckER baseline,the proposed entity enhancement methods improve mean reciprocal rank by 26.34 percent,perform better on low frequency entities,and effectively alleviate the cold start problem.Overall,this study provides a practical route for the automatic acquisition,completion,and explainable reasoning of fault knowledge,which helps improve the efficiency of intelligent operation and maintenance.

参考文献/References:

［1］XU D,CHEN W,PENG W,et al.Large language models for generative information extraction:a survey［J］.Frontiers of Computer Science,2024,18(6):186357.
［2］BEKOULIS G,DELEU J,DEMEESTER T,et al.Joint entity recognition and relation extraction as a multi-head selection problem［J］.Expert Systems with Applications,2018,114:34-45.
［3］WANG Q,MAO Z,WANG B,et al.Knowledge graph embedding:a survey of approaches and applications［J］.IEEE Transactions on Knowledge and Data Engineering,2017,29(12):2724-2743.
［4］ROSSI A,BARBOSA D,FIRMANI D,et al.Knowledge graph embedding for link prediction:a comparative analysis［J］.ACM Transactions on Knowledge Discovery from Data,2021,15(2):1-49.
［5］SOLTANSHAHI M A,TEIMOURPOUR B,ZARE H.BuB:a builder-booster model for link prediction on knowledge graphs［J］.Applied Network Science,2023,8(1):27.
［6］成志鑫,曾跞,唐骁,等.基于知识图谱的校园分布式光伏发电站设备故障自动化检测系统［J］.电子设计工程,2025,33(1):86-90. CHENG Zhixin,ZENG Li,TANG Xiao,et al.Knowledge graph-based automatic fault detection system for equipment in campus distributed photovoltaic power stations［J］.Electronic Design Engineering,2025,33(1):86-90.
［7］戴小建.光伏故障处置知识图谱构建及辅助决策研究［D］.济南:山东建筑大学,2024.
［8］LIANG S,ZHU A,ZHANG J,et al.Hyper-node relational graph attention network for multi-modal knowledge graph completion［J］.ACM Transactions on Multimedia Computing,Communications and Applications,2023,19(2):1-21.
［9］LIU Y,MA Y,HILDEBRANDT M,et al.Tlogic:temporal logical rules for explainable link forecasting on temporal knowledge graphs［C］//Proceedings of the AAAI Conference on Artificial Intelligence,2022,36(4):4120-4127.
［10］TANG K,LI S,TANG J,et al.Fusing structural information with knowledge enhanced text representation for knowledge graph completion［J］.Data Mining and Knowledge Discovery,2024,38(3):1316-1333.
［11］GUO W,LI Z,WANG X,et al.Convd:attention enhanced dynamic convolutional embeddings for knowledge graph completion［J］.IEEE Transactions on Knowledge and Data Engineering,2025,37(9):5049-5062.
［12］徐浩,康振渊,张焱,等.融合知识图谱与大语言模型的储能领域知识推荐研究［J］.南京工程学院学报(社会科学版),2025,25(1):82-90. XU Hao,KANG Zhenyuan,ZHANG Yan,et al.Knowledge recommendation in the energy storage domain integrating knowledge graph and large language model［J］.Journal of Nanjing Institute of Technology(Social Science Edition),2025,25(1):82-90.
［13］徐浩,葛琳琳,张焱,等.面向辅助决策的领域知识图谱构建及其场景式应用研究［J］.科技情报研究,2025,7(4):58-69. XU Hao,GE Linlin,ZHANG Yan,et al.Construction of domain knowledge graph for decision support and its scenario-based applications［J］.Scientific Information Research,2025,7(4):58-69.
［14］SCHWEIKERT J,STUCKY K-U,S W,et al.A photovoltaic system model integrating fair digital objects and ontologies［J］.Energies,2023,16(3):144.
［15］徐浩,康振渊,张焱,等.面向电力变压器故障辅助决策的知识图谱构建及补全策略研究［J］.中国安全生产科学技术,2025,21(5):46-54. XU Hao,KANG Zhenyuan,ZHANG Yan,et al.Knowledge graph construction and completion strategy for auxiliary decision-making of power transformer faults［J］.Journal of Safety Science and Technology,2025,21(5):46-54.
［16］ZHOU H,LI M,XIAO Y,et al.LEAP:LLM instruction-example adaptive prompting framework for biomedical relation extraction［J］.Journal of the American Medical Informatics Association,2024,31(9):2010-2018.
［17］BALAEVIC＇ I,ALLEN C,HOSPEDALES T.Tucker:tensor factorization for knowledge graph completion［C］//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing(EMNLP-IJCNLP),2019:5185-5194.
［18］DUBEY S R,SINGH S K,CHAUDHURI B B.Activation functions in deep learning:a comprehensive survey and benchmark［J］.Neurocomputing,2022,503:92-108.
［19］WANG Z,LIANG J,LI R,et al.An approach to cold-start link prediction:establishing connections between non-topological and topological information［J］.IEEE Transactions on Knowledge and Data Engineering,2016,28(11):2857-2870.

相似文献/References:

[1]张凯,周清,商景林,等.基于大语言模型的生产安全（工矿商贸）事故报告数字化方法*[J].中国安全生产科学技术,2025,21(1):38.[doi:10.11731/j.issn.1673-193x.2025.01.005]
　ZHANG Kai,ZHOU Qing,SHANG Jinglin,et al.Digitization method of work safety (industry,mining,commerce and trade) accident reports based on Large Language Model[J].Journal of Safety Science and Technology,2025,21(3):38.[doi:10.11731/j.issn.1673-193x.2025.01.005]
[2]王明达,赵宝熙,吴志生,等.基于大语言模型的燃气事故调查报告实体识别*[J].中国安全生产科学技术,2025,21(2):139.[doi:10.11731/j.issn.1673-193x.2025.02.018]
　WANG Mingda,ZHAO Baoxi,WU Zhisheng,et al.Entity recognition of gas accident investigation reports based on large language model[J].Journal of Safety Science and Technology,2025,21(3):139.[doi:10.11731/j.issn.1673-193x.2025.02.018]
[3]蔡春城,刘永,宿国瑞,等.基于大语言模型增强的瓦斯事故实体抽取和图谱预警研究*[J].中国安全生产科学技术,2025,21(11):90.[doi:10.11731/j.issn.1673-193x.2025.11.011]
　CAI Chuncheng,LIU Yong,SU Guorui,et al.Research on entity extraction and graph-based early warning of gas accidents enhanced by large language models[J].Journal of Safety Science and Technology,2025,21(3):90.[doi:10.11731/j.issn.1673-193x.2025.11.011]

备注/Memo

备注/Memo:
收稿日期： 2025-12-19
* 基金项目： 国家社会科学基金重点项目（25ATQ008）；国家自然科学基金项目（72502104）；江苏高校哲学社会科学研究重大项目（2024SJZD066）
作者简介： 徐浩，博士，副教授，主要研究方向为信息智能处理与检索。
通信作者： 张焱，博士，教授，主要研究方向为电力系统安全、技术经济分析。
更新日期/Last Update: 2026-03-31