Research On Intelligent Decision-making Methods In Parallel Emergency Management Of A Chemical Cluster

The atmospheric pollution incidents caused by spontaneous or anthropogenic activities in a chemical cluster can exert fatal chemical disasters.However,current management of a chemical cluster accident has some critical defects.For example,people cannot study the management of a chemical cluster accident through field experiment,so that the emergency management plan is decided according to governor's experience and intuition,lacking scientific validation.Futunately,the atmospheric real-time monitoring and intelligent decision methods in a chemical cluster based on parallel emergency management is able to assit governors in disaster warning,risk analyzing and experimental computing,which can reduce the risk of accidents in preparation period of emergency management.The study aims to construct real-time monitoring and intelligent decision models for parallel emergency management.After reviewing the limitations and development of current monitoring big data analysis methods,monitoring network optimization methods and monitoring resources utizalition methods,this dissertation investigates the comprehensive monitoring big data analysis methds,data-driven monitoring network optimization model and data-driven monitoring resources utizalition models in a novel way.Real-time monitoring and intelligent decision in a chemical cluster based on data statistic,Bayesian maximum entropy,optimization methods and game theory are proposed.Especially,some breakthroughs concerning monitoring resources utizalition game theory are made.The contributions and deliverables are concluded as follows:(1)Proposing comprehensive monitoring big data statistical and analyzing method.Other than time series analysis of concentration data,wind field analysis,clustering analysis and source term estimation are conducted in this study.Time series releasing regularity,monsoon regularity,clustering regularity and information of emission sources are obtained at meantime.These regularties are meaningful for instructions on managing monitoring resources in a chemical cluster,but can also be used for subsequent modeling.(2)Proposing a data-driven monitoring optimization method by combining Bayesian maximum entropy and multi-objective optimization model.Based on long-term monitoring measurenments and location information of monitoring resources,the future spatiotemporal concentration trend can be acquired through Bayesian maximum entropy method.Further,a multi-objective optimization model can be used to optimize the layout of monitoring network by importing the predicted data.The method proposed in this study is the first one focusing iterative optimization of monitoring network in the background of a chemical cluster.(3)Constructing a series of chemical plant environmental protection games by combining the source estimation algorithm and the security Stackelberg game.Aiming at the scheduling requirements of various monitoring resources in chemical clusters,this study firstly proposes different game models,which are suitable for different monitoring resources(e.g.,fixed monitoring stations and mobile monitoring vehicles).These game models combined with source estimation algorithms can realize the following functions: the scheduling of monitoring resources;the detecting of emission sources and the verification of emission behaviors,which help governors to effectively manage the chemical plants.(4)Intensifying the application of data-driven modeling technology in the parallel simulation framework and designing the prototype system for parallel emergency management.Based on the dynamic data-driven modeling technology,the monitoring network optimization model and the chemical plant environmental protection games are constructed.By importing the real-time data collected by real-world sensors,the model parameters can be adjusted.This study finally proposes a parallel emergency management framework for a chemical cluster based on a variety of theoretical methods and designs a prototype system with the ability of trend analysis,comprehensive monitoring,forecasting and post warning and dynamic planning.(5)Validating the feasibility of the prototype system via simulation experiments and real data.The study conducts a series of simulation experiments as well as field experiments and applies the actual monitoring big data,environmental data and meteorological data of Shanghai chemical cluster,and finally verifies the validity and practicability of the proposed algorithm,model and prototype system.The focus of this dissertation is the study of real-time monitoring and intelligent decision methods for parallel emergency management.On the basis of enriching the monitoring big data analysis methods,the data-driven air quality monitoring network optimization method and the data-driven chemical plant environmental protection game models are researched innovatively.To enrich the data collecting ways,gas sensor modules and UAV-based monitoring platform are designed.The study can promote the understanding of the emergency management in a chemical cluster,and it is of great significance to assist managers to make decisions and plans.