大语言模型融合知识图谱的风电运维问答系统研究

风电场运维工作高度依赖现场实践经验,而行业内人员高流动性带来经验传递难题,传统知识库和问答系统日益暴露出其局限性。为提高问答系统在专业领域的适用性和可靠性,设计了一种融合大型语言模型和知识图谱的风电运维问答系统,通过语义理解和关联性分析,整合结构化与非结构化数据,提供全面、准确的专业回答。主、客观评价表明,该专业问答模型的准确性、连贯性及信息量均优于某中文大语言模型及ChatGLM模型,不仅提升了风电运维的效率,也为行业知识传承和更新提供了解决方案。

Graphs of NMS Flows on S^3 with Knotted Saddle Orbits and No Heteroclinic Trajectories

We consider NMS systems on S^3 without heteroclinic trajectories connecting two saddles orbits and we build the dual graphs associated with this kind of flows. We prove that flows coming from essential attachments of orientable round 1-handles can be reproduced from the corresponding dual graph.