基于BP神经网络的谢尔曼单位线优选算法研究与实践  

Application of Sherman Unit Hydrograph Optimization Algorithm Based on BP Neural Network

作　　者：王惟一 WANG Weiyi(Liaoning Runzhong Water Supply Co.,Ltd.,Shenyang 110003,Liaoning,China)

机构地区：[1]辽宁润中供水有限责任公司,辽宁沈阳110003

出　　处：《水力发电》2024年第8期5-10,共6页Water Power

摘　　要：洪水预报是防洪减灾的重要技术手段,但传统预报过程受人为经验因素影响较大。将辽宁指数模型、谢尔曼单位线、BP神经网络模型有机结合,构建LNZS-UH-BP模型。以降雨总量、降雨量时段最大、降雨强度、暴雨中心、径流深总量、径流深时段最大、初始蓄水量、单位线洪峰流量、单位线洪峰时段9项指标为输入节点,单位线编号为输出节点,构建9-3-1架构的BP神经网络优选模型,该模型可结合场次降雨特点自动优选单位线进行产汇流计算。以葠窝水库为研究区域,采用历史30场典型洪水资料对模型进行率定及精度验证。结果表明:LNZS-UH-BP模型在葠窝水库得到很好的应用,洪水要素及洪水过程都得到了很好的模拟,既提高了预报精度,又延长了预见期,使洪水预报更准确、更智能。采用LNZS-UH-BP模型进行洪水预报是可行的,在相似地区具有重要的借鉴意义。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.

关 键 词：辽宁指数模型 谢尔曼单位线 BP神经网络 LNZS-UH-BP模型 葠窝水库 

分 类 号：P338[天文地球—水文科学]