Prediction Of Secondary Flood In Semi-Arid Area Based On Multiple Combination Models

The silt dams play an important role in the long-term stability of the Yellow River.Because of the drainage flow of the silt dam is small,in the Short-lasting rainstorm process,there will be a large amount of water interception in the dam,which will bring great threat to the dam system.Therefore,accurate the sub-storm flood forecast has become an important issue for the safety of silt dams.In semi-arid area,the storm flood process is restricted by many factors.During a heavy rain time,rainfall difference,space-time distribution is uneven,rainstorm center is also difficult to find its laws,showing a lot of randomness and the changes of the underlying surface of the basin directly or indirectly affect the relationship between rainstorm and flood.Therefore,the analysis of the influence of each factor on the flood of the storm flood can provide scientific basis for the safety of the silt dam.In view of this,the paper takes the chabagou watershed in the semi-arid area of northern Shaanxi as the research object,carries on the study of the influence factors of the storm flood,constructs the single forecast model and the combination forecast model theory frame,finally evaluates the model prediction effect,and analyzes the applicability of the model.The main research contents are as follows:(1)The change trend of flood characteristic quantity and the influence analysis of climatic factors.By using the Kendall test method and the Spearman test method,this paper analyzes the trend of flood season rainfall,flood season rainfall.The storm flood and the rainstorm series in Chabagou watershed,and analyses its persistence by R/S test.(2)Identification of factors affecting the storm flood.The characteristics of rainfall in the 1980-2010 watershed were analyzed,including:the rainfall,maximum 3-hour rain intensity,centroid of storm rainfall and centroid of storm rainfall.Based on the land use remote sensing data and DEM map provided by the Institute of Remote Sensing Application of the Chinese Academy of Sciences,the underlying factors such as land use and vegetation coverage in the watershed in 1986,1995,2000,2008 and 2013 were extracted.The blocking and storage of the silt dam is the important underlying factor affecting the secondary flood,so it is necessary to restore the silt dam so as to get the secondary flood after the restoration.(3)Selected 20 flood in the Chabagou watershed was simulated,of which 16 were used regularly and four floods were used to verify the flow.multivariate linear regression model,BP neural network model and the deterministic coefficient analysis were built to take the selected flood as input variable and the flood as output variable.The analysis of a single predictive model has a good predictive effect,the combination of the optimal weighted combination model of three kinds of single models,the variable weights based on the error square and the reciprocal variable weight combination model,and the combination model of MLR-DCCA BP combined model based on the internal combination of the model,Compared with the single model,the combined model has higher applicability,in which the BP neural network model based on MLR-DCCA is more stable than single model,fixed weight and variable weight combination model,which can be used to forecast the secondary flood of silt dam system.