| 1,827 | 1 | 196 |
| 下载次数 | 被引频次 | 阅读次数 |
为了解决目前在线大语言模型在水利行业应用中出现的长尾知识、时效性、AI幻觉和垂直领域回答不准确等问题，通过结合检索增强生成(RAG)功能，构建个人知识库，融合DeepSeek本地部署，达到以较小成本解决水利行业中在线大语言模型使用时存在的相关问题。在提出本地布置DeepSeek-RAG问答系统方案的同时，对该方案进行了定性和定量测试。结果表明，基于DeepSeek-RAG技术本地部署水利行业智能问答系统对水利行业的垂直领域的回答有了针对性的显著增强，并对行业内部相关信息的时效性、可解释性有了一定提升，同时在一定程度上解决了在线大语言模型的AI幻觉问题。
Abstract：To address issues such as long-tail knowledge, timeliness, AI hallucinations, and inaccuracies in vertical field responses encountered in the application of online large language model in the water resources industry, this paper constructs a personal knowledge base by combining retrieval augmented generation(RAG), and solves the related problems in the use of online large language model in water resources industry at a small cost by local deployment of DeepSeek. Other than suggesting the local deployment of DeepSeek-RAG-based Q&A system, this paper also conducts qualitative and quantitative tests of the solution. The results show that the local deployment of intelligent Q&A system in water resources industry based on DeepSeek-RAG technology has significantly enhanced the pertinence of the answers in vertical field of water resources industry. The timeliness and interpretability of the relevant information in the industry have been improved. At the same time, it solves the problem of AI hallucinations in online large language models to a certain extent.
[1] 张浩森，徐远翔，万水明.大语言模型赋能的水利知识平台技术研究[J].水利规划与设计，2025(9):100-106.(ZHANG H S,XU Y X,WAN S M.Research on the technology of water conservancy knowledge platform empowered by large language models[J].Water resources planning and design,2025(9):100-106.(in Chinese))
[2] 李书举，任照博，窦立星，等.无线电管理行业DeepSeek AI模型本地化部署中的安全方案浅析[J].中国无线电，2025(4):69-71.(LI S J,REN Z B,DOU L X,et al.A brief analysis of security solutions for the localized deployment of DeepSeek AI models in the radio management industry[J].China radio,2025(4):69-71.(in Chinese))
[3] 洪辉，林俊伟，肖铮，等.DeepSeek驱动的高校图书馆AI应用发展现状、挑战及前景——与云瀚联盟专家的对话与思考[J].信息与管理研究，2025,10(3):32-40.(HONG H,LIN J W,XIAO Z,et al.Current status,challenges,and prospects of DeepSeek-driven AI applications in university libraries:Expert dialogues and analysis from the Yunhan community[J].Journal of information and management,2025,10(3):32-40.(in Chinese))
[4] 张志鑫，明晨曦，刘颉，等.基于JRAG的涉水法律法规智能知识问答技术[J].人民长江，2025,56(2):240-247.(ZHANG Z X,MING C X,LIU J,et al.Research on intelligent knowledge Q & A for water related laws and regulations based on jointly retrieval-augmented generation[J].Yangtze River,2025,56(2):240-247.(in Chinese))
[5] 中华人民共和国水利部.洪水影响评价技术导则：SL/T 808—2025[S].北京：中国水利水电出版社，2025.(Ministry of Water Resources,the People’s Republic of China.Technical guideline for flood impact assessment:SL/T 808—2025[S].Beijing:China Water&Power Press,2025.(in Chinese))
[6] 何果.基于大模型的洪涝灾害防御数字化孪生系统的研究与实现[D].西安：西安理工大学，2024.(HE G.Research and implementation of digital twin system for flood disaster prevention based on large language model[D].Xi’an:Xi’an University of Technology,2024.(in Chinese))
[7] 马乐平，易善桢，严冬，等.基于WebGIS的灌区水资源管理信息化方法与实现[J].水电能源科学，2010,28(8):137-139.(MA L P,YI S Z,YAN D,et al.WebGIS-based water resources management information system and its implementation in irrigation area[J].Water resources and power,2010,28(8):137-139.(in Chinese))
[8] PENG B C,ZHU Y,LIU Y C,et al.Graph retrieval-augmented generation:A survey[J].ACM transactions on information systems,2026,44(2):1-52.
[9] 黄健辉.大模型在水利行业网络安全防护中的应用[J].网络安全和信息化，2024(10):129-131.(HUANG J H.Application of large model in network security protection of water conservancy industry[J].Cybersecurity & informatization,2024(10):129-131.(in Chinese)
基本信息:
DOI：10.20040/j.cnki.1000-7709.2026.20251406
中图分类号:TP18;TP391.1;TV21
引用信息:
[1]程功,王洁,果利娟,等.基于DeepSeek-RAG技术的水利行业智能问答系统研究[J].水电能源科学,2026,44(06):242-246.DOI:10.20040/j.cnki.1000-7709.2026.20251406.
2025-08-11
2025
2026-04-20
2025-08-23
2025
1
2026-01-08
2026-01-08
2026-01-08