【作者】 邹强；

【导师】 周建中；

【作者基本信息】 华中科技大学 ， 系统分析与集成， 2013， 博士

【摘要】 中国是一个洪涝灾害频繁发生的国家，并且以其发生频率高、分布范围广、灾害程度重、经济损失大而成为最为严重的自然灾害，严重威胁人民生命财产安全和阻碍了社会经济可持续发展。因此，开展洪灾风险分析研究，确切分析洪灾发生的可能性及其可能造成的损失，对于科学、有效地防洪减灾具有重要意义。然而，洪水灾害系统是一个包括致灾因子、孕灾环境、承灾体和灾情的开放式复杂巨系统，具有高度非线性、时空动态性和复杂不确定性，系统中各种耦合问题日趋复杂并呈现由低维向高维的演化过程。毫无疑问，针对这一复杂系统的风险分析与风险管理，无论是采用经典的控制理论，还是采用传统的数学手段，都将遇到诸如信息不确定性、模型合理可靠性、数据资料不完备性、系统集成普适性等方面的困难，相关研究已呈现出“从低维线性到复杂高维非线性”、“从单尺度到多维时空尺度”、“从单情景到组合情景”、“从单变量到多重耦合变量”、“从确定性到不确定性”的发展趋势，迫切需要理论与方法体系的创新，亟待研究和发展新的理论与方法体系。本文基于复杂耦合的思想和复杂系统的集成方法，提出了多种复杂耦合评价模型并应用到不确定性洪灾风险综合分析与智能评价的研究中，旨在丰富和发展洪灾风险分析中的指标体系构建、综合权重推求、评价方法建立的技术路线，实现对复杂条件下洪灾风险识别、诊断、评价以及不完备信息条件下洪灾风险模糊评价，为洪水风险综合管理和防灾减灾紧急预案提供科学的决策支持。具体说来，本文的主要研究集中在洪水危险性识别、洪灾易损性诊断、洪灾风险综合评价、模糊风险评价和洪灾灾情智能评价等方面，取得了一些具有理论意义和实用价值的研究成果。本文的主要研究内容及创新成果如下：（1）针对洪水分类指标时空分布不均、交叉严重且综合权重不易计算等问题，采用灵敏度系数将主、客观权重综合为组合权重，提出了一种基于组合权重的模糊聚类迭代模型；同时，对基本差分进化算法进行改进提出了性能更优的自适应混沌差分进化算法（ACDE），全局搜索得到了组合权重模糊聚类迭代模型的最优的模糊聚类中心和灵敏度系数。实例研究表明提出的分类方法合理、可靠、稳健，既能有效处理洪水分类指标的不确定性和模糊性，又综合考虑了评价指标的主客观权重，在无评价标准情况的排序、评价和决策问题中具有良好的推广应用价值。（2）针对洪灾易损评价指标体系的不相容性和不确定性，以多属性决策的理论与方法为基础，将理想解法和灰色关联法相结合并进行有效改进，对研究区域在发生洪水时可能造成的损失程度进行动态综合评价。研究表明，提出的改进灰色理想解法，思路清晰、结果合理、应用灵活，从位置和形状相似性上综合描述了样本与理想解的接近程度，能充分利用知识信息，更好地挖掘数据的内在规律，提高了洪灾易损性评价的科学性，在多指标综合评价方面具有较好的应用前景。（3）依据灾害系统理论，针对洪水灾害形成、发展、致灾过程不确定性对洪灾风险的影响，结合水文气象、社会经济、自然环境等数据统计资料以及相应的法规、案例，构建了了荆江分洪区洪灾风险综合评价指标体系及其等级评价标准，并运用梯形模糊数和层次分析法相结合的方法（TrFN-AHP）确定评价指标权重。同时，针对复杂洪灾系统中随机、模糊、灰色等各种不确定性，以洪灾风险管理广义熵智能分析为理论框架，提出了基于最大熵原理和属性区间识别理论的洪灾风险综合评价模型（AIRM-POME），最大程度地消除洪灾风险分析中的各种不确定性，由置信度准则和特征值公式判定各评价单元的洪灾危险等级和易损等级，并根据自然灾害风险的定义及其定量表达式给出风险等级。实例研究表明评价结果可信度高，是一种洪灾风险分析的新方法，可推广应用到其他自然灾害的风险分析中。（4）传统概率计算方法在实际应用中会遭遇“小样本”的不完备信息问题，存在不精确和不确定的理论瓶颈。为有效处理洪灾风险的不确定性和不完备性，在以“信息分配”和“信息扩散”为核心的模糊信息优化处理技术的基础上，应用改进的内集-外集模型得到了洪灾可能性-概率风险值，既阐明了在现有不完备条件下超越概率估算不确定、不精确、不唯一的客观事实，也为风险评价中的模糊信息识别、容纳、处理与计算提供了一条新思路；在此基础上，计算了区间风险估计值，得到了与软直方图基本一致的结果，并引入模糊集截集技术推求得到了不同截集水平下的冒险风险值、保守风险值和最大可能风险值，体现了风险评价结果的不同层次结构，可为决策者提供多层次、多值化的风险信息。（5）针对洪灾灾情评估指标与灾情等级之间的复杂非线性关系，引入支持向量回归建模方法并结合ACDE进行参数优选，同时将集对分析理论、三角模糊数、随机模拟技术进行结合，将离散的整数型灾情等级评价推广到连续型、区间型灾情等级综合评价，解析了灾害评估指标体系与实际灾情的正相关映射规律，克服了传统评价方法合理构造评价指标集与评价等级之间函数关系的困难，攻克了评估结果易出现振荡性误差的技术瓶颈，突破了洪灾灾情等级高精度、高分辨率动态综合评价的理论障碍，丰富和发展了多级模糊条件下洪水灾情综合评价的理论与方法体系。更多还原

【Abstract】 Flood disasters are among the most frequent and devastating types of disasters over theworld. Worldwide statistics indicate that continuously increasing flood damages and lossesof human lives remain at high levels.Therefore, it is necessary to analyze flood risk toensure healthy and sustainable economic development, and flood risk assessment hasbecome worldwide one of the hot issues in the field of natural science and technology.And the flood disaster system is an open complex giant System with highly nonlinear,dynamic and complex uncertainties, including the disaster-inducing factors,disaster-breeding environment, hazards-bearing bodies and flood disaster loss, and theproblmes in the system become increasingly complex and coupled problems fromlow-dimensional to high-dimensional. Undoubtedly, according to risk analysis and riskmanagement of this complex system, regardless the classical control theory or traditionalmathematical methods, we would likely encounter much trouble, such as whetherinformation is uncertainty, whether the models are reasonable and reliable, whether thedata is complete and the integrated system is universal. Moreover, development trend ofrelated research has rendering out "from low dimension linear to complex high dimensionnonlinear","from single scale to multidimensional spatio-temporal scale","from singlestories to combination stories","from single variable to multiple coupled variable" and"from certainty to uncertainty", so urgent need is to propose and develop new theories andmethodologies. Based on coupling complicated idea and complex system integrationapproach, a variety of complex coupling evaluation model were proposed in this articleand then applied to the study of flood risk comprehensive analysis and intelligentevaluation under uncertainty. The aim of this article is to enrich and develop technicalroute of the establishment of flood risk analysis index system, determination of integratedweight and establishment of evaluation methods, then to reach the realization of flood riskidentification, diagnostics, evaluation, and incomplete information under complexconditions as well as fuzzy risk analysis under incomplete condition, hence to providescientific decision support for flood risk management and disaster prevention andmitigation integrated emergency plans. Specifically, the main research focus of thisarcticle are flood hazard identification, flood vulnerability diagnosis, flood riskcomprehensive assessment, flood risk fuzzy analysis and flood disaster intelligentevaluation, and the main research contents and innovations are as follows:(1) For the flood classification indices are uneven distribution, cross-serious and theircomprehensive weight are hard to calculate and so on, the sensitivity coefficient was usedto combine the subjective and objective weights into integrated weights, then fuzzy clustering iterative model with integrated weights was proposed for flood classfication.Then adaptive chaotic differential evolution algorithm (ACDE) was proposed for globalsearch of the optimal fuzzy clustering center and sensitivity coefficient in the combinedweight fuzzy clustering iterative model. Case study shows that the proposed classificationmethod is reasonable, reliable, robust, which can effectively deal with flood classificationindexes under uncertainty and ambiguity, and has good application value in sorting,evaluation and decision-making problems with no evaluation criteria.(2) For there are incompatibilities and uncertainties in the flood vulnerabilityassessment index system, based on multi-attribute decision making theory and method, wecombine and improve the Technique for Order Preference by Similarity to Ideal Solution(TOPSIS) and Grey Correlation method (GC) into IGC-TOPSIS for dynamiccomprehensive evaluation of disaster loss caused by flooding. Research shows thatIGC-TOPSIS is reasonable and flexible, and it is able to describe the closeness from theposition and shape of the similarity to the ideal solution and make full use of knowledgeand information, to better tap the inherent data law, thus to improve the scientific of floodvulnerability assessment, and have has good application prospects in multi-indexcomprehensive evaluation.(3) Based on disaster system theory and taking consideration into flood formation,development and process on flood risk, as well as combined with hydrometeorological,socio-economic and natural environment statistics data and corresponding regulations andcases, the flood risk evaluation index system of flood diversion district as well as itsevaluation standards were established. Then the analytic hierarchy process combined withtrapezoidal fuzzy numbers (TrFN-AHP) was employed to determine the weights forevaluation indices. Meanwhile, considering the stochastic, fuzzy, gray and otheruncertainties in flood disaster risk system and based on the intelligent analysis ofgeneralized entropy theoretical framework, we proposed the flood risk comprehensiveassessment model using the princple of maximum entropy and attribute intervalrecognition theory (AIRM-POME), to maximize the elimination of flood risk analysis in avariety of uncertainties, and the confidence criterion and eigenvalues equation are adoptedto determine the flood hazard and flood vulnerability grades. Finally, according to thedefinition and quantification expressions of risk, we obtain the flood risk grades for eachunit. Case study shows the proposed method has high reliability evaluation result, and it isa new method for flood risk comprehensive analysis and can be extended to other naturaldisasters risk analysis.(4) Traditional probability calculation method will encounter "small sample" incomplete information in practical applications, and there exists imprecise and uncertaintheories bottleneck. In order to effectively deal with the uncertainty and incompleteness offlood risk, based on "information distribution" and "information diffusion" as the core ofthe fuzzy information optimization processing technology, we employ the improvedinterior-outer-set model (IIOSM) to obtain the possibility-probability distribution (PPD)risk results, which could not only clarify the objective facts of existing incompleteexceedance probability estimation under uncertainty and imprecision, but also provide anew idea for risk assessment to identify, accommodate, process and compute the fuzzyinformation. on this basis, the interval risk estimates were calculated, which wasconsistent the results by soft histogram. And furthermore, by combining PPD with fuzzyset cut technique, we got the conrresponding venture risk, conservative risk and maximumpossibility risk values under different levels, thus can provide decision makers withmulti-level, multi-valued risk information.(5) Considering the complex nonlinear relationship between flood disaster assessmentindices and disaster grades, we adopt the support vector machine combined with ACDEfor parameter optimization for flood disaster evaluation, and then set pair analysis theory,triangular fuzzy number, stochastic simulation technique are combined, thus to extend thediscrete integer disaster grade to continuous and interval disaster grade. With the twomethods above, we are able to analysis the related mapping rules between disasterevaluation model and actual disaster, overcome the difficulty of the traditional evaluationmethods in establishing functional relationship between assessment indices and grades,overcome the technical bottleneck of calculating assessment results with oscillatorydeviation, break the theoretical obstaclesthe of flood disaster grade high-precision andhigh-resolution dynamic comprehensive evaluation, as well as enrich and develop thetheory and method of flood disaster loss comprehensive assessment under multi-levelfuzzy conditons.更多还原