基于大数据的雾霾天气情景构建与应急决策鲁棒优化方法研究

项目名称： 基于大数据的雾霾天气情景构建与应急决策鲁棒优化方法研究

项目编号： No.71473146

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 管理科学

项目作者： 黄弘

作者单位： 清华大学

项目金额： 60万元

中文摘要： 重度雾霾事件频繁发生，对城市居民的安全和健康造成了严重的危害。雾霾已成为目前最亟待应对的重大公共危机事件之一。本课题针对雾霾这一重大公共危机事件，开展基于大数据的雾霾预测和应急决策优化研究。研究雾霾天气与城市多种时空动态数据（气象、交通、排放源、人群流动、网络媒体等）的时空相关性，探索雾霾与城市物理维度和社会维度特性的时空交互影响规律；研究雾霾天气的大数据挖掘方法与模型，开展基于大数据的雾霾预测与情景构建；研究雾霾天气应急策略的目标评价数学模型，构建雾霾天气的多目标应急鲁棒决策优化流程与方法；在此基础上，选取北京市等典型城市开展实证性研究。通过本研究，可以加深理解雾霾特性规律，深化和拓展大数据分析与应用，丰富管理科学内涵，进而为政府高效应对雾霾事件提供科学支撑。

中文关键词： 雾霾；大数据；情景构建；鲁棒优化；多目标决策

英文摘要： The frequently occurring severe haze has severely threatened the citizen's safety and health, and becomes one of the most important public crisis issues requiring the urgent research and response. In this project, we study the prediction of haze based on big data and develop the optimization method for emergency decision making of haze. The spatial and temporal correlations among the haze and the multiple urban dynamic data (e.g. meteorology, traffic, emission source, human mobility and network media, et al.) will be analyzed. The interaction effects between the urban multiple dynamic characteristics and the haze in both physical and social fields will be explored. The method and models for big data mining of haze will be developed for the prediction and construction of haze scenarios. The mathematical model for the objective evaluation of haze weather emergency management strategies will be proposed, and the multi-objective optimization procedure and method for robust emergency decision making will be developed. Based on the above research, typical cities such as Beijing will be selected for the application studies. This project will help us well understand the characteristics of haze, deepen and broaden the application of big data analysis and application, enrich the connotation of management science, and then provide scientific basis for the government's efficient response to the haze weather.

英文关键词： Haze;Big Data;Scenario Construction;Robust Optimization;Multi-objective Decision Making