基于大语言模型和知识图谱的数字孪生流域知识平台设计  

Design of Digital Twin Basin Knowledge Platform based on Large Language Model and Knowledge Graph

作　　者：李巍[1,2] 李树元[2] LI Wei;LI Shu-yuan(Haihe River Water Conservancy Commission,MWR,Tianjin 300170,China;Tianjin Lonwin Technology Co.,Ltd.,Tianjin 300170,China)

机构地区：[1]水利部海河水利委员会,天津300170 [2]天津市龙网科技发展有限公司,天津300170

出　　处：《海河水利》2023年第12期75-78,共4页Haihe Water Resources

摘　　要：设计了一款基于生成式预训练大语言模型的数字孪生永定河流域知识平台,该平台基于海量的业务数据、图书文献、期刊论文和行业资料,构建数字孪生永定河知识库。以知识库为基础,训练大语言模型,创建水利对象、预报方案、调度预案、历史场景等知识图谱。基于大语言模型、知识图谱和向量数据库,实现知识检索和智能问答,并开发水利对象关联查询、业务规则匹配、历史场景复演、实时调度方案编制等水利业务智能应用。A digital twin knowledge platform for the Yongding River Basin based on generative pre-trained large language models is designed.The platform builds a digital twin knowledge base for the Yongding River based on massive business data,books,periodicals,and other materials.Based on the knowledge base,the large language model was trained to create knowledge graphs of water conservancy objects,forecasting schemes,scheduling plans,historical scenarios,and so on.Based on the large language model,knowledge graph,and vector database,knowledge retrieval and intelligent question answering were realized,and intelligent applications for water conservancy business such as intelligent retrieval of water conservancy object information,business rule matching,historical scenario replay,and real-time scheduling plan preparation were developed.

关 键 词：大语言模型 知识平台 知识图谱 数字孪生流域 

分 类 号：TP399[自动化与计算机技术—计算机应用技术] TV882.8[自动化与计算机技术—计算机科学与技术]