基于多智能体的洪涝风险动态评估理论模型

随着全球气候变化的加剧和城市化进程的快速发展,洪涝灾害的影响和复杂性日益加剧。洪涝灾害风险评估是一项复杂的系统工程,在防洪减灾工程理论与实践中具有重要意义;作为洪涝灾害风险管理和应急处置的基础和核心,风险评估模型直接影响洪涝灾害风险评估结果的可靠性。洪涝灾害风险是洪涝灾害复杂系统组成部分相互作用的结果,综合考虑洪涝灾害系统的特点以及智能体建模的优势,提出了一种基于多智能体的洪涝风险动态评估理论模型。首先从复杂系统建模的角度出发,对洪涝灾害复杂系统进行分析建模,构建基于多智能体的洪涝灾害风险动态评估框架模型;其次,对所构建的框架模型中单个智能体模型进行研究,分别建立基于反应型的孕灾环境智能体模型和基于慎思型的致灾因子、承灾体和风险分析智能体模型,并对以上单个智能体的流程进行研究;然后以风险分析智能体联盟为例对模型框架中所涉及到4类智能体联盟的内部结构和流程进行分析,对构建的框架模型中涉及到的智能体通信协调和作用规则进行探讨;最后,在Netlogo平台上,基于30 m DEM数据和构建的模型,对淮河流域的暴雨型洪涝孕育发生发展全过程的人口风险进行动态评估。结果表明,构建的模型能有效评估洪涝灾害全过程中人口风险的动态。研究结果对洪涝灾害人口灾情评估、应急救助和应急管理都具有较强的指导意义。

A multi-agent based theoretical model for dynamic flood disaster risk assessment

As the consequences of urbanization trends and climate changes in recent years, the impacts and interactions of flood disasters have become significantly complicated. As a complex system engineering, flood disaster risk assessment takes on very important meaning in the theory and practice of flood control. What is more, the risk assessment model is the foundation and core of risk management and emergency response of flood disaster. So, the flood disaster assessment model has a deterministic influence on the flood disasters risk assessment results. The flood disaster risk is the consequence of the interactions between different factors in a complicated flood disaster system. According to the characteristics of flood disaster complex system, agent-based modeling (ABM) technology, belonging to complex system modeling, is employed in dynamic risk assessment for flood disaster. This paper proposed a multi-agent based theoretical model for dynamic flood risk assessment. First, the framework of dynamic risk assessment model, namely flood risk dynamical assessment multi agent system (FRDAMAS), is constructed for flood disaster based on the analysis of the complex flood disaster system from the perspective of complex system modeling. Second, based on research on each kind of agent model in FRDAMAS, a reactive agent model is used to construct single agent model of hazard environment and a deliberative agent model is used to construct the agent models of hazard factor, hazard effect object and risk analysis. Meanwhile, the procedures, the mechanism and the rules of abovementioned agents are studied. Third, taking risk analysis agents alliance as an example, we analyze interior structures and procedures of the four types of agents alliance involved in FRDAMAS, including hazard environment agents alliance, hazard factor agents alliance, hazard effect object agents alliance, and risk analysis agents alliance. Finally, the dynamic assessment of risk population during the whole process of the rain-storm flood disaster in the Huaihe River Basin is implemented using the proposed method and DEM data with a spatial resolution of 30 m on NetLogo platform. The results demonstrate that the proposed method can effectively assess the dynamic changes of population risk during the whole process of flood disaster. And it is also important that the results will have a strong guiding significance for population risk assessment, emergence response, and emergence management for flood disasters.