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  

A knowledge-driven approach for intelligent generation of hydraulic engineering contingency plans:A case study of the Middle Route of South-to-North Water Diversion Project

作　　者：刘雪梅 卢汉康 李海瑞 槐先锋 陈晓璐 LIU Xuemei;LU Hankang;LI Hairui;HUAI Xianfeng;CHEN Xiaolu(School of Information Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450000,China;Collaborative Innovation Center for Efficient Utilization of Water Resources,Zhengzhou 450000,China;School of Management and Economics,North China University of Water Resources and Electric Power,Zhengzhou 450000,China;China South-to-North Water Diversion Group Middle Line Co.,Ltd,Beijing 100038,China)

机构地区：[1]华北水利水电大学信息工程学院,河南郑州450000 [2]黄河流域水资源高效利用省部共建协同创新中心,河南郑州450000 [3]华北水利水电大学管理与经济学院,河南郑州450000 [4]中国南水北调集团中线有限公司,北京100038

出　　处：《水利学报》2023年第6期666-676,共11页Journal of Hydraulic Engineering

基　　金：河南省科学院科技开放合作项目(220901008);国家自然科学基金项目(72271091)。

摘　　要：水利工程传统应急方案存在数字化程度低、内容关联性差、智能辅助决策不足等问题。本文利用知识图谱和深度学习技术,创建一种水利工程应急方案智能生成模式。首先基于风险防控手册及险情抢险应急方案等文本,提出应急方案知识图谱本体模型,构建应急方案知识图谱,实现应急方案文本中非结构信息的结构化表达。其次,基于水利工程巡检文本,利用BERT(Bi-directional Encoder Representation from Transformers)+BiLSTM+CRF(Bi-directional Long Short Term Memory with Conditional Random Fields)实体识别模型,智能识别巡检文本中的风险事件、工程等实体。最后,设计应急方案智能生成模板,通过多特征融合的实体对齐技术、知识检索与推理技术,实现应急方案的智能生成与推送。通过模型准确性分析以及“渠道渗漏”等实例验证,本文方法识别准确率高(F 1值为96.21%),生成的应急方案可靠,可推广到水利工程应急抢险以及应急预案智能生成等应急管理工作中。Traditional emergency plans for hydraulic engineering projects have problems such as low digitisation,poor content relevance and insufficient intelligent aid for decision-making.In this paper,we use the knowledge graph and deep learning technology to create an intelligent generation model of emergency plans for hydraulic engineering projects.Firstly,based on the text of risk prevention and control manual and the emergency plan,we propose a knowledge graph ontology model of emergency plan,construct a knowledge graph of emergency plan,and realise the structured expression of non-structural information in the text of emergency plan.Secondly,based on the water resources engineering inspection text,we use BERT(Bi-directional Encoder Representation from Transformers)+BiLSTM+CRF(Bi-directional Long Short Term Memory with Conditional Random Fields)entity recognition model to intelligently identify risk events,projects and other entities in the inspection text.Finally,an intelligent generation template for emergency solutions is designed,and through multi-feature fusion of entity alignment technology,knowledge retrieval and inference technology,the intelligent generation and pushing of emergency solutions is realised.Through the model accuracy analysis and the validation of examples such as“channel leakage”,this paper shows a high recognition accuracy(F 1 value of 96.21%)and a reliable emergency plan generation,which can be extended to emergency management such as emergency rescue of hydraulic engineering projects and intelligent generation of emergency plans.

关 键 词：知识图谱 水利工程 应急方案 南水北调 

分 类 号：X93[环境科学与工程—安全科学] TV513[水利工程—水利水电工程]