【作者】 王富强；

【导师】 许士国；

【作者基本信息】 大连理工大学 ， 水文学及水资源， 2008， 博士

【摘要】 中长期水文预报不仅在水库调度、防洪减灾等工作中有重要的作用,而且在洪水资源利用、水权管理等方面也有很重要的意义,因此,中长期水文预报一直是水文工作者深入探讨的课题。近年来,随着计算机技术的发展和新的数学方法的不断涌现,中长期水文预报得到了较快的发展。但是,由于其复杂性和数据资料等因素的制约,中长期水文预报研究仍处在发展阶段,相对于短期水文预报来说,滞后于生产实际的要求。在预报理论研究上,更多注重的是水文系列的统计相关特性,而对物理成因关系关注的相对较少;在预报方法上,对各种方法的有效性研究不够,使现有的方法很难在实践中推广应用;在预报结果的实际应用上,中长期水文预报目前主要是对水资源的宏观调控起一些参考性作用。基于此,本文探讨了基于物理因子分析的中长期水文预报方法,并将预报成果用于指导平原河流洪水资源利用工作。主要研究内容和成果概述如下:(1)从水文循环的机理出发,综合分析影响区域水文情势的物理因素,主要包括天文因素、海表温度、以及大气环流等。详细分析了太阳黑子活动情况、日月地三球位置关系、北太平洋海温冷暖变化、ENSO事件以及大气环流因子等物理因素对区域水文情势的影响,为进行基于物理因子分析的中长期水文预报方法研究提供资料准备和理论支持。(2)针对现有水旱灾害趋势预测方法无法体现未来洪水发生可能性和量级的缺陷,根据气象因素与水旱灾害的关系,引入随机过程的概念,提出了区域水旱趋势预测的转移概率、太阳活动相位、厄尔尼诺事件等三种方法,推导了相应的计算公式,综合三种方法预测结果得出最终结论。结合东北区水旱灾害史料,分析了其多年来的水旱灾害特征,并对该区2001～2010年水早趋势进行了预测,不仅为水早灾害时域特性的研究探索了一条新的途径,也可在一定程度上为区域洪水资源利用长期规划的制定、洪水风险管理等工作提供有益参考。(3)针对中长期水文预报数据资料的数量和种类繁多,数据间关系复杂,难于检索有用信息并组织用于预报的问题,将关联规则数据挖掘分析方法引入到中长期水文预报研究中,研究了中长期径流关联规则模式的提取及预测方法。首先,结合中长期水文预报的特殊性,根据预报目标初选物理影响因子。然后,根据关联规则挖掘算法的要求对数据进行清洗和预处理,构成预报事务数据集。最后,面对预报事务数据集进行关联规则挖掘,提取满足事先设定的最小支持度和最小置信度的强关联规则,解释规则并建立模型进行预测。实例分析证明,该方法在保证一定精度的情况下大大减小了工作量,有助于从海量数据中提取对预报目标有意义的关联规则和模式。(4)针对水文中长期预报中单一的定量预报方法精度偏低,稳定性差,不能满足实际生产活动的要求这一情况,提出了定性定量嵌套的多因子神经网络预报模型。利用人工神经网络灵活多变的拓扑结构和强大的非线性逼近能力,基于前期物理影响因子分析作为输入量,通过改变输出节点的个数,先建立定性预报模型,然后在定性预报的基础上建立定量预报模型,最后综合定性预报和定量预报的结果得出结论。实例分析证明,定性定量嵌套的多因子神经网络预报模型不但有一定的物理成因基础,而且可以较好地克服传统使用单一定量预报模型进行预报的盲目性,提高了预报的精度和可靠性。(5)在洪水资源利用中,存在的各种未来信息的不确定性是风险的重要来源。中长期预报可以在不同程度上对洪水资源利用系统中各类因素的未来状态进行界定,从而确定面向整个时期引蓄洪水的时机及其后期安全性。本章以白城为例,研究了中长期水文预报在平原河流洪水资源利用中的应用。基于中长期预报成果,并结合考虑时间因素与蓄水状态的河流洪水资源利用二维风险分析模型,以改变各蓄水单元蓄水状态可能带来的后期损失作为主要的风险损失,兼顾时间与蓄水状态两维特性,对洪水资源利用效益和风险损失进行计算,从而得到考虑中长期预报信息的汛期不同时段各蓄水单元引蓄不同洪量的风险率,为洪水资源利用预案制定和实施提供依据。2005、2006年的洪水资源利用实例证明,较高精度的中长期预报成果对该地区洪水资源利用预案制定和实施起到了积极的作用,可最大限度的避减洪水资源利用中的风险和损失,实现最大效益。最后,对全文进行了总结,并对有待进一步研究的问题进行了展望。更多还原

【Abstract】 Medium and long-term hydrological forecasting plays an important role ranging from reservoir operation, flood control yield to disaster mitigation, floodwater resource utilization, water right management, etc. Medium and long-term hydrological forecasting has been researched by many hydrology scientists. With the new techniques of computer and mathematics methods coming out, medium and long-term hydrological forecasting has a fast development. But it is still under development and lagged behind production process compared to short-term forecast because of its complexity and short of hydrologic data. More attention has been paid to statistical properties in hydrological series in the prediction theory study, and less in physics causes analysis. Neglecting the validity of forecasting methods makes the existing methods not effectively popularized and applied, and used only as a reference in macro-control of water resource in practical application. For these reasons, this paper studied on methods of long-term hydrological forecasting based on physics factor analysis and its application in plain flood resource utilization. The main contents and results are as follows:(1) Physical factors of regional hydrologic influence are analyzed from the mechanism of hydrological cycle, including astronomical factors, sea surface temperature, atmospheric circulation factors, and so on. Sunspot activity, relations among sun, moon and earth, changes of North Pacific sea surface temperature, ENSO events, atmospheric circulation factors which affect regional hydrological situation are analyzed in this paper. It gives a theoretical support for the study of medium and long-term hydrological forecasting based on physical factors analysis.(2) The relationship between the meteorological factors and the flood hazards has been established, considering that the existing forecasting methods can not express the possibility and degree of flood. Three methods for forecasting the trends of floods and droughts are proposed, and the mathematical formulae are derived through the combination of the meteorological and stochastic concepts and methods. These three methods are the shift probability, the phase of the Sun activities, El Nino events. These methods have been used to forecast the future trends of floods and droughts in Northeast China, which have been proved to be a new way to the study of flood hazard time domain characteristics and an effective reference for guiding regional flood utilization and the practical flood risk management. (3) Association rules mining is introduced to medium and long-term hydrological forecasting, aiming on the variety and complex relationships of long-term hydrological forecasting data and difficulty search and management for forecasting. The model of long-term hydrological forecasting based on association rules mining is established in this paper. Firstly, forecast factors are selected to constitute the long-term forecast database based on the forecast object considering the characteristics of hydrology forecast. Secondly, the forecast database is established based on data cleaning and pretreatment. Finally, the strong association rules which accord with the min-support and min-confidence are extracted by the method of apriori. Long-term hydrological forecasting model is established based on the strong association rules. A case is studied to validate this new model, and the results reveal that the model can reduce workload and is helpful to finding the interesting strong association rules for the goal of forecasting from massive data.(4) Long-term qualitative and quantitative synthesis prediction model has been established based on the variable topological structure of artificial neural network in this paper, because that the quantitative forecasting method has large errors and the qualitative forecasting methods can not satisfy the production process, Firstly, the qualitative forecasting model is established based on physical factors analysis. Secondly, the quantitative forecasting model is established based on changing output node. Finally, conclusions are drawn by combining qualitative and quantitative forecasting results. Two cases are used to validate this new method, and the results reveal that the method is a better way to improve forecasting accuracy and stability than either of the models used separately.(5) There are different kinds of future information uncertainty in the flood resource utilization. They are the main resources of risk. The future state about all kinds of input information in the system can be defined in different degree by using the results of medium and long-term hydrological prediction. The paper analyzes the application of long-term runoff forecast in plain flood resource utilization. Therefore, the risk and the loss during the process of flood utilization can be avoided or reduced. The preplan of flood utilization can be made and the risk and the loss during the process of flood utilization can be avoided or reduced as more as possible based on the results of medium-term runoff forecasting. The application examples of 2005 and 2006 show that the higher precision of long-term hydrological forecast play a positive effort in making and actualizing flood resource utilization preplan. The risk and the loss during the process of flood utilization has been avoided and decreased to the greatest extent and the flood utilization has achieved the biggest benefit by applying the medium and long-term hydrological prediction.Finally, conclusions are made, and problems for further study are reviewed.更多还原