Mining and Forecasting of Rainstorm Disaster Chain Based on Knowledge Graph
In recent years,under the influence of climate change and urbanization,extreme rainstorm events in urbanized areas have been increasing in China.Rainstorm disasters not only trigger secondary disasters such as landslides and urban flooding,but also result in a series of derivative consequences such as transportation interruptions,power outages,and damage to municipal facilities.This phenomenon,where rainstorms serve as direct or indirect factors inducing a series of secondary and derivative disasters to occur consecutively,is called a rainstorm disaster chain.It often exhibits cumulative and unpredictable characteristics,making its evolution and development difficult to predict and control.With the development of big data technology,it is possible to transform massive data into knowledge,providing a new approach for risk prediction of rainstorm disaster chain.This paper aims to explore the evolution process of rainstorm disaster chain and predict the secondary disasters and derivative consequences,and thereby prevent the risk of rainstorm disaster chain effectively.In this paper,a knowledge graph for rainstorm disaster chain is constructed,incorporating the impact on vulnerable entities into the conceptual layer of rainstorm disaster chain ontology model.Then,a rainstorm disaster chain mining method is proposed based on link coupling of knowledge graph.By mapping the knowledge graph to Bayesian network structure,rainstorm disaster chain prediction model based on Bayesian networks is constructed.Finally,focusing on the Pearl River Delta region as the study area,various types of rainstorm disaster chains are explored,and the secondary disasters and derivative consequences are predicted by taking the"8.29 Shenzhen Rainstorm"event as an example.The research results indicate that:(1)The knowledge graph constructed based on rainstorm disaster chain ontology model can not only depict the secondary disaster but also illustrate the derivative consequences such as infrastructure damage,casualties,and potential impacts on livelihoods and production caused by rainstorms.The rainstorm disasters in the Pearl River Delta region are often accompanied by lightning disasters and strong wind disasters,which can easily lead to secondary disasters such as floods,waterlogging,debris flows.They are also prone to causing casualties,collapse of house walls,damage to municipal facilities,interruption of transportation.(2)The proposed rainstorm disaster chain mining method based on knowledge graph link coupling mines the entire rainstorm disaster chain by coupling multiple independent events.It reveals the evolutionary mechanism of rainstorm disaster chain,consisting of"rainstorm disaster-secondary disasters-derivative consequences".The main types of rainstorm disaster chains in the Pearl River Delta include natural disasters,casualties,urban lifeline system disruption,equipment and construction accidents,and potential impacts on production and life.(3)The proposed rainstorm disaster chain prediction model based on Bayesian networks can accurately predict the secondary disasters and derivative consequences of rainstorm disaster chains.Using the August 2018 rainstorm event in Shenzhen as an example,the model accurately predicted the most probable disaster chain at the early stage of the rainstorm disaster event,with results consistent with the actual situation,providing decision-making basis for disaster mitigation and risk reduction of rainstorm disaster chain.The framework of rainstorm disaster chain ontology constructed by incorporating the consequences on vulnerable entities can mine the completed rainstorm disaster chain,and reveal the secondary-derivative mechanism of rainstorm disaster chain.The proposed Bayesian network prediction model of rainstorm disaster chain based on knowledge graph improves the accuracy and reliability of rainstorm disaster chain prediction.The Bayesian network based on the mined rainstorm disaster chain overcomes the subjectivity of traditional expert methods,and the knowledge graph integrates all the information of rainstorm disaster events effectively,providing a comprehensive data basis for Bayesian network prediction.
万方数据
维普
