| With the development of flood forecasting science,hydrological models have been widely used to solve problems including hydrology and water resources,social and huaman development such as environmental and ecological.The wet areas of the south are rich in the resources of rain and flood,and the methods of calculating the hydrological model of the basin are more and more mature.But for the half dry area,which occupies 52% of our territory,the work of hydrological models is less,and the use of watershed hydrological models is problematic both at home and abroad.The double super model is especially apply to semi-humid and semi-arid regions,as the preferred model of flood forecasting,should be given full attention.Now the study of the sensitivity and the rate of the model parameters is poor,that cannot provide accurate model information for governments and flood prevention departments.It is difficult to meet the requirements of parameters prediction.Based on double super model as the research object,through analysis of model parameter,identify important influenceparameters of the model output response,reduce the blindness in the choice of parameters,this paper also establishes the system of the classification parameter rate of flood classification in the Shanxi Province,provides the basis for the parameter rate,and improves the reliability and forecast accuracy of the model's operation.In this paper,take the three water stations in Shanxi Province as the research object,including Yushe,shangjing and Loufan.Select the typical flood events in evey research basin for double super model parameter sensitivity analysis.First of all,using local analysis method get the sensitivity and correlation of double supermodel parameters in different river basins,different levels of flood and multiple target functions,and use the variation coefficient method determines the synthetic sensitivity coefficient of the model parameter.Then use the modified LH-OAT method,through directional change on the parameters,the double super model parameters sensitivity and correlation condition is obtained in the different basin,different levels and under the multiple target function,and based on entropy method the synthetic sensitivity coefficient of model parameters determined.Finally,compare and analysis the results,research shows that:(1)The sort for model parameters sensitivity are obtained by local analysis is Sr>b>?0>Ks>??C,The parameters Sr?Ks?b??0 i s the sensitivity parameter,C?? is insensitive parameters.The sort for model parameters sensitivity are obtained by global process is K s>b>Sr>?0>??C,The parameters Sr?Ks?b is the sensitivity param eter,the parameter ?0 is the more sensitive paramete C?? is insensi tive parameters.Comparing the results of the two studies shows that the sensitivity and size of the model parameters are different.But for parametric sensitivity classification,only ?0 is affected by the analysis method,and the other parameter sensitivity grade are stabil ity.(2)Use the local analysis and global process method analysis the correlation between the parameters and the target function,can known that:The correlation between sensitivity parameters and target function Wi?Qmi is clear in different levels of flood and watershed.The parameter?0 ? b is positive correlationwith Wi ? Qmi,the parameters Sr ? Ks are negative correlation with Wi?Qmi.But the parameters are not explicitly relevant to all target functions,and the correlation is not clear when the target function becomes IVF?RE?RSS?PE.This paper use the fuzzy ISODATA iteration model to analyze the historical flood.The flood peak,flow and flood totals of the flood waters are the main targets of flood forecasting,so the flood peak discharge and flood total of selected historical flood are analyzed by cluster analysis.The historical flood are divide into three types,Large flood,medium floods and small floods.Due to the flooding phenomenonis complicated,difficult to control,law of runoff are also different in the different types of floods,to reduce the flood forecast error by using only a set of hydrological forecast model parameters on the whole basin,this paper established the idea of classification rate of hydrological forecast model parameter,to find the same type of bus hong.The classification rate of the hydrological prediction model parameters results shows that:(1)The BP neural network classification model established in this paper can accurately judge the type of flood,and the accuracy is 100% in the sample forecast.(2)In this paper,the established river basin flood classification forecasting methods,increased the percent of pass of the flood volume from 73% to 82%,the relative error of the flood volume reduced from 18.1% to 11.3%;The percent of pass of flood peak also increased from 73% to 82%,while the relative error declined from 16.4% to 14.6%.The forecast accuracy of the research basin is improved,and the reliable basis for the real-time operation of the basin is provided.The traditional perturbation analysis method is used to classify the sensibility of the model parameters,and there is no quantitative calculation of the sensitivity coefficients of model parameters.This article empowered to the target function,and analysised the model synthesis sensitivity coefficient,the results are more comprehensive and reliable.It is of great practical significance to understand the flow mechanism of double super model,reduce the process of model rate and improve the simulation precision of model.The BP neural network classification model,which is established in this paper,can be more accurate in determining the size of flood waters.The proposed scheme of the classification parameters of flood water in the basin which based on the BP neural network classification results,improved the forecast accuracy of the river basin.In addition,the influence of flood classification and identification results on the selection of flood classification characteristics is an issue that should be further studied. |
PDF Full Text Request