【作者】 陈玥；

【导师】 王乘； 黎育红；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2010， 博士

【摘要】 随着我国社会经济的迅速发展和城市化进程的不断加快,洪水灾害已经成为制约经济社会可持续发展、威胁人民生命财产安全保障、以及阻碍我国和谐社会发展进程的重要影响因素。根据洪水灾害系统的不确定性,以及系统的高维性、复杂性、开放性以及动态性,从系统科学的观点看,洪水灾害系统是一个动态复杂大系统。毫无疑问,针对这一复杂大系统的控制与管理,无论是采用经典的控制理论,还是采用传统的运筹学技术,都将遇到诸如数据资料不完备、信息不确定性考虑不充分、模型可靠性低等方面的困难。本文基于定性与定量的综合集成方法,把系统科学的理论和方法以及基于灰色理论和云模型等反精确技术引入对不确定性洪水灾害系统的研究中,系统地探讨了洪水灾害的模拟、预测、评估的综合分析方法,旨在建立洪水灾害分析的不确定性系统分析理论、反精确方法与技术体系,实现对复杂环境下洪水灾害不确定性系统的分析、预测、推理及评价,为进一步对洪水灾害的有效调控和管理提供科学的依据,使社会经济朝着协调的方向发展。具体说来,文章基于灰色理论和云模型的反精确方法,对不确定性洪水灾害系统进行分析研究,包括灾变预测、中长期径流预报、生命损失推理,以及洪灾等级综合评估。其主要研究内容和成果如下：(1)梳理了洪水灾害系统的复杂性特征和基本研究对象,通过对现实世界中的不确定性原理及其普遍存在性的分析,提出对不确定性洪水灾害系统进行分析的反精确性软计算方法的研究方案与具体实现思路。(2)由于实际运用中洪水灾害数据少、不确定性因素影响众多,作者根据灰色理论对序列预测所要求的最低信息量,建立洪水灾变点预测的灰色动态模型群,来对洪水灾害进行基于年径流量的参考性灰色灾变预测。以新疆雅马渡水文站的多年径流量数据为例,对基于灰色模型群的洪水灾变预测进行实例研究。(3)基于对洪水灾害系统的灰色不确定性分析,建立多元时变灰色预测模型,并基于信息熵原理,用熵权法将该模型与非时变的免疫神经网络模型、最小二乘支持向量机模型进行并联组合建模,来做基于年径流量的洪水参考性中长期预报,分析各单项模型各自的优、缺点,以及组合预测建模的实用意义。最后,以新疆伊犁河雅马渡水文站的年径流预测为例,对该站年径流量进行基于信息熵的并联组合预测建模,并与三个单项模型的预测结果进行比较分析,证实了组合预测的合理性、普适性和可靠性。(4)针对国外计算生命损失经验模型的局限性,具体分析导致生命损失的各种影响因素的不确定性,选取起决定作用的三个因素,即洪水强度、预警时间、人们对洪水危害性意识程度,合理划分构成的15种洪水灾害情景的组合模式,按照国内外经验确定洪水灾害生命损失率推理规则,运用表达信息不确定性的云模型来将其定性概念量化建模,然后对定量的待估样本按该规则做定性的推理,得到定性、定量的综合推理。(5)为了克服传统白化权函数的局限,将表示信息模糊性和随机性的定性、定量转换的云模型引入灰色白化权函数的表达,对传统白化权函数进行改进,建立基于灰色云模型的白化权函数,用灰色云聚类模型来对洪水灾害损失进行等级评估。并以1989～1990年间我国部分省市发生的45个洪水灾害的灾情案例作为评估实例,对待估样本进行基于变权、定权的灰色云聚类,得到灾情评估结果,证实了该方法的实用性。更多还原

【Abstract】 With the rapid development of China’s economy and the accelerating process of urbanization, the flood disaster has constrained the economic and social sustainable development, threatened the life and property security, and impeded the process of harmonious social development. According to the uncertainty, the hi-dimensional, complex, open and dynamic characteristics of the flood disaster system, from a system science point of view, flood disaster system is a dynamic complex system. There is no doubt that for the control and management of this complex system, whether classical control theory, or traditional operations research techniques, will encounter difficulties. In this article, based on the comprehensive integration of qualitative and quantitative methods, the system science theories and the anti-accuracy technique based on gray theory and cloud model etc. are introduced in the uncertainty study of flood disaster system. In addition, the integrated analysis approach for the simulation, prediction and evaluation of flood disaster has been systematically explored, for the establishment of uncertainty system theory, anti-accuracy method and technology system for flood disaster analysis. This provides the scientific basis for effective control and management of flood disaster in order to harmoniously develop social economics.Specifically, the article based on the anti-accuracy method of the gray theory and cloud model for analysis of the uncertainty of the flood disaster systems, including disaster prediction, mid- and long-term runoff forecasting, reasoning of life loss, and comprehensive assessment of disaster level. The main research contents and results are as follows:(1) Sorted out the complexity of the flood disaster system and the basic research objective, and analyzed the uncertainty principle and the widespread nature in the real world. On this basis, the anti-accuracy calculation method for analysis of the uncertainty of flood disasters was proposed.(2) Established the gray dynamic models to predict flood disasters on the basis of annual runoff. Taking the Yamadu station in Xinjiang as the research area, the flood disaster was predicted.(3) Based on the gray uncertainty discussion of flood disaster system, a prediction model with time-varying multiple parameters had been set up. Based on the information entropy principle, the entropy-weight was used to parallel combine this model with the non-time-varying immune neural network model and the least-squares support vector machine model for mid- and long-term flood forecasting, analysis of the advantages and disadvantages of each single model and the practical significance of combined predictive modeling. Finally, taking the annual runoff forecasting of Yamadu Station in Xinjiang as a case study, the rationality, universality and reliability of the combined model have been confirmed.(4) Selected three major factors of life loss in flood disasters, including flood intensity, warning time, people’s awareness of flood danger, categorized into 15 flood disaster scenarios of different combination. In accordance with domestic and foreign experience, the inference rule of life loss was determined. The cloud model was used to set up a quantitative model for the qualitative concept, and then with this rule the qualitative reasoning was performed for quantitative samples to be estimated, leading to the comprehensive qualitative and quantitative reasoning results(5) Refined the traditional whitening-weight function using the cloud model, which is a representation of both fuzziness and stochastic, and could take a bi-directional conversion between qualitative and quantitative. In this way, a grey-cloud clustering model is proposed for the synthetic ranking evaluation of flood disaster. The novel method is applied to evaluate the 45 flood disaster cases of several provinces in 1989～1999, and taking the amount of building collapse, affected area, number of victims, direct economic loss as evaluation indicators. The result shows that this method has much rationality and utility for flood disaster evaluation.更多还原