基于BP神经网络的谢尔曼单位线优选算法研究与实践OACSTPCD

Flood forecasting is an important technical means for flood control and disaster reduction,but the traditional forecasting process is greatly influenced by human experience factors.This article combines the Liaoning index model,Sherman unit hydrograph,and BP neural network model together to construct the LNZS-UH-BP model.The 9-3-1 BP neural network optimization model is constructed by taking the total rainfall,the maximum rainfall period,the rainfall intensity,the rainstorm center,the total runoff depth,the maximum runoff depth period,the initial storage capacity,the unit line peak flow and the unit line peak flow period as the input nodes,and the unit line number as the output node.The model can automatically select the unit line to calculate the production and confluence according to the characteristics of rainfalls.Taking the Shenwo Reservoir as the research area,the model is calibrated and verified for prediction accuracy using 30 typical historical flood data.Research shows that the LNZS-UH-BP model is well applied in the Shenwo Reservoir,and the flood elements and processes are well simulated.The model not only improves the prediction accuracy,but also extends the foresight period,which making the flood forecasting more accurate and intelligent.It indicates that using the LNZS-UH-BP model for flood forecasting is feasible and has important reference significance in similar areas.