【作者】 周玉良；

【导师】 金菊良；

【作者基本信息】 合肥工业大学 ， 水文学及水资源， 2006， 硕士

【摘要】 水是一种特殊的资源,支撑着所有的生命,是实现可持续发展的重要物质基础。近年来随着我国经济的快速发展和人口迅速增长,水资源短缺、水环境恶化、干旱和洪涝灾害等水安全问题日趋尖锐。 首先,介绍水安全问题研究现状,及智能非参数方法在水安全系统工程中的应用进展。接着,针对水安全系统输入和输出之间的复杂关系,用传统的参数模型描述存在着模型结构和参数难以确定的问题,提出变结构遗传算法、遗传程序设计、人工神经网络、投影寻踪插值、核函数相似插值等智能非参数建模方法。最后给出这些非参数模型在水安全系统中实际应用:(1)在预测方面,对作为水资源系统常用输入的降雨建立了降雨序列自回归预测模型和降雨空间回归预测模型;(2)在评价方面,建立了水质和洪水灾情变结构遗传算法、遗传程序设计、BP神经网络、基于投影寻踪插值、基于核函数相似插值等非参数评价模型;(3)模拟方面,建立了径流序列模拟的半参数模型,其中确定性成分采用参数模型描述,随机性成分采用非参数模型描述。为了能够综合利用现有各种预测和评价模型的有效信息,采用基于人工神经网络的非线性组合预测模型来进行预测和评价。上述智能非参数模型在水安全系统工程的实例应用,均取得了较为合理和有效的结果,表现出一定的可行性和可靠性具有一定的推广应用价值。更多还原

【Abstract】 Water as a special resource, which sustains all life, is the substance of sustainable development society. With the rapid development of economy and growth of population recently, the problem of water resources shortage、 water environment deterioration、 drought and flood becomes more and more acute.Firstly introduce the review of study on water security problem and the research on application of nonparametric method to water security system engineering. Then present intelligence nonparametric modeling method ,such as: changeable structure genetic algorithm(CSGA) method、 genetic programming method(GP)、 artificial neural networks(ANN) model、 interpolation model based on project pursuit(PPIM)、 similitude interpolation model based on kernel function(KFIM) and so on, due to the difficulty of describing the complicated relation between input and output data of water security system by traditional parametric model, in which model structure and parameter is difficult of being determined. Finally give the application of those model mentioned above to water security system engineering: (1)in prediction, using rainfall, which is often used as the input of water resources system, as the exploration object, establish the auto regressive rainfall time serial prediction model and the relation between rainfall and space model;(2)in evaluation: using water quality and flood disaster loss as evaluation object, establish the CSGA、 GP、 ANN、 PPIM、 KFEM evaluation model;(3)in simulation: establish semiparametric runoff time serial simulation model, in which the determinacy component is described by parametric method and the indeterminacy component is described by nonparametric method, in order to take advantage of various information from single model, a nonlinear combination prediction and evaluation model based on BP neural networks is applied to forecast rainfall and evaluate flood disaster loss. The application results indicate that: the possibility and reliability of the models applied to water security system engineering are obtained. It is concluded that the models stated above can be applied to water security system engineering generally.更多还原