| 158 | 0 | 25 |
| 下载次数 | 被引频次 | 阅读次数 |
随着“天空地水工”一体化感知体系全面建设，引调水工程安全管理数据呈现出多源异构、规模庞大、动态变化复杂特征，传统的基于单模态数据分析、挖掘方法在工程安全智慧监管场景下面临明显的局限性。融合多模态大模型与知识图谱技术，提出一种“感知—认知—决策”的智能监管范式。基于标准规范、风险应急管理资料、巡检文本及图像、多光谱遥感影像，微调多模态大模型并结合动态提示策略，构建面向工程安全的多模态知识图谱；利用检索增强生成技术及知识图谱的结构化知识，提升大模型在专业领域的可靠性及推理能力；提出多智能体协同的决策链构建方法，通过动态任务编排实现模型能力耦合，赋能工程安全管理中的风险识别、评估及预案生成业务。实验结果表明，本研究方法的多模态知识提取准确性高，可支撑引调水工程安全智慧监管。
Abstract：With the comprehensive development of the “sky-space-earth-water-project” integrated monitoring system, safety management data of water diversion projects exhibit characteristics of multi-source heterogeneity, large volume, and dynamic complexity. Traditional analysis and mining methods based on single-modality data face significant limitations in the context of intelligent safety supervision. By integrating multimodal large models with knowledge graph technology, an intelligent supervision paradigm of “perception–cognition–decision” is proposed. Based on standards and specifications, risk and emergency management materials, inspection texts and images, and multispectral remote sensing imagery, a multimodal large model is fine-tuned and combined with a dynamic prompting strategy to construct a multimodal knowledge graph for engineering safety. Retrieval augmented generation(RAG) and the structured knowledge within the knowledge graph are employed to enhance the model's reliability and reasoning capability in specialized domains. A collaborative multi-agent decision chain construction method is introduced, enabling the coupling of model capabilities through dynamic task orchestration to support risk identification, assessment, and contingency planning in safety management. Experimental results show that the proposed method achieves high accuracy in multimodal knowledge extraction, providing effective support for intelligent safety supervision of water diversion projects.
[1]李国英.进一步全面深化水利改革为推动水利高质量发展、保障我国水安全作出新的贡献[J].中国水利，2025(2):1-9.
[2]李勇.南水北调工程运行管理成效与推进后续工程高质量发展的思考[J].中国水利，2024(20):6-9+14.
[3]高长胜，胡江.气候变化对我国水利基础设施安全影响及应对建议[J].中国水利，2024(24):60-67.
[4]WANG L,LIU X,LIU Y,etal.Multimodal knowledge graph construction for risk identification in water diversion projects[J].Journal of Hydrology,2024,635:131155.
[5]WANG L,LIU X,DONG Y,et al.Risk Evolution Analysis of Cross-Regional Water Diversion Projects Based on Spatio-Temporal Knowledge Graphs[J].Journal of Hydrology,2025,650:132533.
[6]刘雪梅，卢汉康，李海瑞，等.知识驱动的水利工程应急方案智能生成方法--以南水北调中线工程为例[J].水利学报，2023,54(6):666-676.
[7]刘业森，赵英虎，张丽伟，等.天津市防洪工程调度知识平台构建与实践[J].中国水利，2025(4):48-55.
[8]LUO T,WU C,DUAN L.Fishbone diagram and risk matrix analysis method and its application in safety assessment of natural gas spherical tank[J].Journal of Cleaner Production,2018,174:296-304.
[9]HOSSEINI S,IVANOV D.Bayesian networks for supply chain risk,resilience and ripple effect analysis:A literature review[J].Expert systems with applications,2020,161:113649.
[10]XU Z,DANG Y.Automated digital cause-andeffect diagrams to assist causal analysis in problemsolving:a data-driven approach[J].International Journal of Production Research,2020,58(17):5359-5379.
[11]吴一戎.遥感技术赋能智慧水利：问题、挑战与建议[J].中国水利，2024(11):1-8.
[12]何健，肖尧轩，刘国庆，等.流域防洪“四预”关键技术研究与应用[J].中国水利，2025(7):14-29.
[13]张锐，张双虎，王本德，等.考虑上游溃坝洪水的水库漫坝失事模糊风险分析[J].水利学报，2016,47(4):509-517.
[14]刘家宏，周晋军，王浩.梯级水电枢纽群巨灾风险分析与防控研究综述[J].水利学报，2023,54(1):34-44.
[15]孙亮，王瑞国，袁瑞，等.人工智能技术在智慧水利中的应用与展望[J].中国水利，2024(3):44-51.
[16]梁辉，赵文光，郭胜山，等.高拱坝-地基体系整体稳定概率地震风险分析[J].水利学报，2021,52(3):310-322.
[17]MAZUMDER R K,SALMAN A M,Li Y.Failure risk analysis of pipelines using data-driven machine learning algorithms[J].Structural safety,2021,89:102047.
[18]LIU A,FENG B,XUE B,et al.Deepseek-v3technical report[J].arXiv:2412.19437,2024.
[19]LI Y,STARLY B.Building a knowledge graph to enrich ChatGPT responses in manufacturing service discovery[J].Journal of Industrial Information Integration,2024(40):100612.
[20]GLM T,ZENG A,XU B,et al.Chatglm:A family of large language models from glm-130b to glm-4all tools[J].arXiv:2406.12793,2024.
[21]LUO Y,ZHANG J,FAN S,et al.Biomedgpt:Open multimodal generative pre-trained transformer for biomedicine[J].arXiv:2308.09442,2023.
[22]钱峰，成建国，夏润亮，等.水利大模型的建设思路、构建框架与应用场景初探[J].中国水利，2024(9):9-19.
[23]舒全英，马媛，陈亮，等.数字孪生水利建设中的人工智能大模型应用探索[J].中国水利，2025(6):14-30.
[24]杨阳蕊，潘世峰，刘雪梅，等.多模态知识图谱与大模型协同的水利工程风险应对决策推荐[J/OL].水利学报，1-12[2025-05-02].https://doi.org/10.13243/j.cnki.slxb.20240575.
[25]杨阳蕊，朱亚萍，陈思思，等.融合群体智能策略的AI链在大坝防汛抢险知识推理中的应用[J].水利学报，2023,54(9):1122-1132.
[26]张天鸿，王晓玲，余红玲，等.基于大语言模型的灌浆工程知识服务系统[J].水利学报，2025,56(1):130-142.
[27]周逸凡，段浩，赵红莉，等.水文模型知识图谱构建与应用[J].水利学报，2024,55(1):80-91.
[28]ZHANG M,WANG J.Automatic extraction of flooding control knowledge from rich literature texts using deep learning[J].Applied Sciences,2023,13(4):2115.
[29]QIU Q,TIAN M,HUANG Z,et al.Chinese engineering geological named entity recognition by fusing multi-features and data enhancement using deep learning[J].Expert Systems with Applications,2024,238:121925.
[30]杨阳蕊，朱亚萍，刘雪梅，等.水利工程文本中抢险实体和关系的智能分析与提取[J].水利学报，2023,54(7):818-828.
基本信息:
DOI：
中图分类号:TV68
引用信息:
[1]王立虎,刘雪梅,李海瑞等.引调水工程安全智慧监管多模态大模型构建技术研究[J].中国水利,2025,No.1013(11):8-19.
基金信息:
国家重点研发计划（2024YFC3210802）; 国家自然科学基金项目（72271091）; 2022年度水利部重大科技项目（SKS-2022029）; 河南省科学院科技开放合作项目（220901008）; 河南省高等教育重点研发项目（24A520021）