Research On Flood Forecasting Of The Upper Weihe River Based On Artificial Neural Network

Flood control is a kind of knowledge prevention and preventing from flood loss, therefore the research is great significance of river flood control.The generating of floods is complicated, so traditional flood forecasting methods need large amounts of data, it could not be predicted using the traditional forecasting methods in the case of material insufficiency. This paper introducts the BP artificial neural network based on the nonlinear the spread of river flood characteristics and establishes a new river flood forecasting model.The author chose hydrological-north word Wushan hydrological inlets of the upstream of Weihe river. The main content of the study are as follows:(1) It briefly expounds the main method of current flood forecasting, introducing the BP nerve network model, and introduces the research status of BP neural network in the river flood forecast. Using MATLAB (2009a) as the model realize platform and water level and hydrological traffic data of 1991-2001, training and test the BP neural network model and saving the results better network, then forecast the flow of the flood in quantitative using the flood of August 2002 as the forecast samples.(2) According to the difference between choosing the number of Hydrological stations, the research is divided into three kinds of models, so as to select the most suitable prediction model of this research.Through the basic BP neural network model for northern road station flood flow forecast, the results of finalize the design coefficients were 0.37,0.89,0.95. The forecast results prove the BP neural network model could be used for this section flood flow forecast.(3) According to the manifests in basic BP neural network model in the prediction, putting forward some basic BP network model-GA-the improved model BPLM, GA-BPBR neural network, through the two improved model to predict A and B model, and make analysis of the three kinds of neural network predictive, getting GA-BPBR neural network prediction results can achieve A precision, whether in the training time or predicting the results have proved GA-BPBR neural network model could forecast river flood flow study area quickly and accurately.(4) Finally, model GA-BPBR is used to extended forecast period river of flood forecasting in the study area, the result is also acceptable.It is proved the BP neural network in the river flood forecast is feasible for the study of the area in a timely and accurate forecast of river flood flow and it provides new ideas and methods of correction, and would play a certain role in making corresponding flood control policy-decision.