泾河年径流量BP神经网络模型研究

Research on BP Neural Network Model for Runoff in Jinghe River Basin

以1957—2000年的实测降水序列和泾河年径流量序列为研究对象,利用EMD法和GA建立了泾河年径流量的BP神经网络模型。分析结果表明:泾河流域年降水量变化可能存在准2～3、5～7、10～13、18～22 a的周期;基于EMD的年径流量BP神经网络模型预测值的相对误差为-4.71%～8.21%,基于GA的年径流量BP神经网络模型预测值的相对误差为2.25%～12.22%。

An annual precipitation-runoff forecasting BP neural network model based on EMD and GA was established,which using the rainfall data from 1957 to 2000 of 4 typical hydrologic stations in Jinghe River basin and the annual runoff as a research object.The results show that the precipitation series has periods that about 2-3,5-7,10-13 and 18-22 years.And using the BP neural network model based on EMD to predict annual runoff,the relative error of predictive value between 4.71% and 8.21%,while using BP neural network model based on genetic algorithm,the relative error of predictive value between 2.25% and 12.22%.