【作者】 李亚娇；

【导师】 李怀恩；

【作者基本信息】 西安理工大学 ， 环境工程， 2003， 硕士

【摘要】 洪涝灾害在我国从古至今都是十分突出的，水库泥沙淤积对水库的不利影响在我国黄土高原地区也是非常严重的，因此对于水库入库洪水预报与出库含沙量预测的研究有着重要的意义。 本文结合黑河流域的特点针对水库的入库洪水预报进行研究，同时探讨了泥沙淤积较为严重的冯家山水库的出库含沙量预测问题，力求能够根据现有实测数据对水库的入库洪水过程和出库含沙量过程进行较为可靠的预测。本文的主要内容和结果如下： 1．结合黑河流域黑峪口以上控制面积较小、地处半湿润地区的特性，采用集总的新安江模型进行黑河流域的洪水预报，效果很好。 2．采用P～Pa～R三变量相关图进行产流计算，然后用总径流响应函数求出该流域净雨的响应函数，从而进行水情预报，效果较好，但稍逊于新安江模型。 3．以当前时段以前n个时段的连续降雨作为输入，采用一个隐层的BP网络对黑河流域洪水预报进行尝试性研究，训练部分效果较好，但网络的泛化能力不高。 4．同BP网络的输入，在RBF网络的输入层与隐层中加入一个与无因次单位线纵坐标相同的权值，以此对黑河流域洪水预报进行尝试性研究，训练部分效果较好，但网络的泛化能力不高。 5．对传统水文模型与人工神经网络在洪水预报中的应用进行比较、分析各模型的优劣及神经网络泛化能力较低的产生原因，并对神经网络应用于黑河流域洪水预报的改进提出建议。 6．采用RBF网络对冯家山水库出库含沙量预测进行尝试性研究，根据洪水入库时间与开闸排沙时间的不同分别选择网络结构，预测结果确定性系数较大，效果 西安理工大学硕士学位论文较好。更多还原

【Abstract】 Dameges of flood are always severe in China in all ages, deposition of sediment in reservoir are serious in Loess Plateau, too. So forecast of inflow hydrograph and out flow sediment concentration of reservoir are very important.According to the characteristics of Heihe watershed, flood hydrograph into Jinpen reservoir is studied. At the same time, prediction of outflow sediment concentration of Feng Jia Shan reservoir is modeled too. The objective is to get more exactly prediction of flood hydrograph and sediment concentration profile based on measured data. The main contents and results of this paper are as follows:1. Considering the characteristics of Heihe watershed that has smaller area and belongs to semi-humid region, Xinanjiang Model is adopted to predict inflow of Jinpen reservoir and the results are satisfied.2. P~Pa~R Correlation Diagram is adopted to calculate net rain, then the response function model is used to predict flood hydrograph, the results are not as good as those of Xinanjiang Model.3.With the input of duration rainfall series before current time, BP ANN model with one hidden layer is tried to predict flood. The results are good for the training data set, but generalisation is weaker.4.The input is the same as that of BP model, weights with the same value of dimensionless unit hydrograph are been added to RBF ANN model between input layer and hidden layer to predict flood. The results are good in the training data set, but generalisation is weaker.5.Traditional hydrologic models are compared with ANN models in floodprediction, avail of every model and cause of weaker generalisation of ANN is analyzed, at the same time advices on improving ANN in flood prediction are proposed.6. RBF model is tried to predict outflow sediment concentration of Feng Jia Shan reservoir. Based on time of flood entering reservoir and the time of opening brake, the network structure is chosen. The results are good.更多还原