突发重大自然灾害下人群疏散及人员伤亡评估研究-基于GIS和多智能体建模

Reducing casualties is the core objective of emergency management, this paper is mainly focus on two vulnerable question of crowd evacuation behavior research at home, like lack of basic data and verification of relationship between behavior and casualties. Firstly, establish the basic data of crowd evacuation behavior by developing complete random sampling field investigation of affected crowds in 4 earthquake and 2 tornado cases in our country. Secondly, clarify the spatial and temporal characteristics and key influencing factors of peoples’ evacuation behavior in sudden major natural disasters, and put forward systematic border, connotation, subjects and conceptual model by spatio-temporal data analysis and statistical methods based on Geographic Information System (GIS) technology. Thirdly, establish GIS-based multi-agent model of crowd evacuation and related casualties in sudden natural disasters and make dynamic simulation of multi-agent interaction and reveal crowd evacuation complex behavior rule by depending on AnyLogic software, basing on the empirical and modeling research of 4 earthquakes and 2 tornadoes in the early stage of our research group, also integrated application of GIS and multi-agent (MAS) theory. Finally, clarify the quantitative relationship between crowd evacuation and its related casualties by implement multiple simulation experiment setup of 3 major kinds of experimental groups (disorder spontaneous evacuation, orderly spontaneous evacuation, organized emergency evacuation) through expert advice, brain storm. Research in this paper is planned to put forward countermeasures and suggestions for emergency management of crowd evacuation with the purpose of reducing casualties in two types of sudden nature disasters by exploring the Multi-Agent Modeling for Crowd Evacuation and related casualties.

减少人员伤亡是突发事件应急管理工作的核心目标，本研究聚焦我国突发自然灾害下人群疏散研究“缺疏散行为数据”“人群疏散与人员伤亡量化关系少模型验证”2个核心问题，通过开展4个地震和2个龙卷风受灾区域人群完全随机抽样调查，建立疏散行为基础数据库；基于时空数据分析，厘清突发自然灾害下人群疏散行为时空特征和关键影响因素，提出人群疏散系统的边界、内涵和概念模型；以课题组4次地震和2次龙卷风实证及建模研究为基础，综合地理信息技术（GIS）和多智能体（MAS）理论，建立基于GIS的人群疏散及人员伤亡MAS模型，动态模拟自然灾害下多主体间交互作用，揭示人群疏散个体和社会行为规律；设置“无序自发疏散、有序自发疏散、有组织应急疏散”3类实验组，进行多重仿真实验，明确疏散行为与人员伤亡的量化关系。研究拟通过对人群疏散MAS模型进行有益探索，提出2类突发自然灾害下人群疏散应急管理对策建议，达到减少人员伤亡的目的。

我国重大自然灾害频繁发生，大多难以预测且破坏力极强，持续威胁公众生命财产安全，突发事件应急管理研究面临巨大挑战。疏散是灾害突发时受灾人群面临的首要问题，也是应急管理的核心任务，高效的疏散行为对于减少人员伤亡具有显著意义。人群疏散是典型多主体复杂系统，摸清个体行为特征与系统演化规律，模拟可能出现的人员伤亡情况，具有实际应用价值。本项目基于国内外文献资料收集，课题组前期15年在自然灾害应急医学救援领域的研究基础，开展了如下研究工作：①对地震多发地区（四川省6个城市）和龙卷风受灾区域（江苏盐城）人群开展在线调查，获得地震灾区样本量4543例，龙卷风灾区样本量529例，建立了人群疏散行为基础数据库。②构建了不同伤情和伤势的地震和龙卷风两类自然灾害伤员本底数据库，归纳了人员伤亡主要影响要素；进一步厘清了人群疏散行为时空特征和关键影响因素，提出人群疏散系统的边界和内涵。③以课题组4次地震和3次龙卷风实证及建模研究为基础，综合地理信息技术和多智能体理论，优化了自然灾害发生演化模型、建筑物损毁模型、人员分布模型、人员伤亡概率模型，进而依托AnyLogic建模工具，建立了基于GIS的人群疏散及人员伤亡多智能体模型，在宏观层面，可模拟地震和龙卷风两类自然灾害下人员疏散与人员伤亡结局；在微观层面，可动态模拟地震灾害下单体建筑内人群疏散时交互作用，揭示人群疏散个体和社会行为特点；从灾害、人群、应急管理3方面入手，设置“自然灾害预警”“自然灾害发生时间”“应急疏散状态”“人群安全认知”等实验组，进行多重仿真“干预”实验，明确了越早发布灾害预警信息、应急管理介入的有组织应急疏散、人群掌握应急疏散知识和流程能够有效降低自然灾害下人员的伤亡率，研究能够为防灾减灾提供循证依据。