BP_Adaboost模型和分组建模的新安江模型在大伙房水库控制流域洪水预报中的应用

Application of BP_Adaboost Model and Xin’anjiang Model of Grouped Modeling in Flood Forecasting of DHF Reservoir Basin

采用BPAdaboost模型和分组建模的新安江模型对大伙房水库控制流域(简称大伙房流域)洪水进行预报,探讨了不同模型方法在丰、平、枯3种代表年洪水中的预报精度。结果表明,分组建模的新安江模型和BPAdaboost模型均适用于大伙房流域的洪水预报,二者预报精度等级均为甲等,洪水过程确定性系数均为乙等。其中新安江模型分组率定参数进行建模预报的方法是可行的,其洪水预报过程与实测过程拟合度较高,能较好地反映洪水的涨落起伏;BPAdaboost模型不仅可以较好地解决传统BP模型预报结果震荡的问题,还能提高预报精度,其模型自修正及参数自动率定功能可以显著提高工作效率,因而在实际应用中可根据需求选择不同模型与方法进行洪水预报。

The Xin’anjiang model of grouped modeling and BPAdaboost model were used to forecast flood of the basin of Dahuofang reservoir(hereinafter referred to as DHF basin),and the forecasting accuracy of these model methods in abundant,flat and dry annual flood was discussed.The results show that both Xin,anjiang model of grouped modeling and BPAdaboost model are suitable for flood forecasting in DHF basin,the forecasting accuracy grade of both models is grade A,and the certainty coefficient of flood process is grade B.Therefore,it is feasible for Xin,anjiang model to modeling and forecast by grouped calibrating parameters,the flood forecasting process fits well with the measured processwhich can better reflect the flood fluctuation.BPAdaboost model can not only solve the problem of oscillation of traditional BP model,but also improve the prediction accuracy,and its function of model self-correction and automatic parameter calibration can significantly improve work efficiency.Therefore,different models and methods can be selected to forecast flood in practical application according to the demand.