基于改进BP算法的地下水动态预测模型

运用学习率自适应动量BP算法建立了吉林西部地下水埋深人工神经网络模拟预测模型。首先利用自回归分析方法确定网络输入输出样本,而后应用“试错法”确定隐含层节点数,最终建立了6∶10∶1的ANN地下水动态模拟预报模型,最后应用VB语言依据改进BP算法编制计算程序进行模拟计算。通过对模型检验可知该模型模拟和预测精度均较高,完全可应用于地下水位动态预报。2002年以后的预报结果表明该地区地下水位持续下降,应及时加以控制。

Dynamic prediction model of groundwater level based on improved BP algorithm

A groundwater depth predication model of artificial neural network(ANN) for West Jilin was established based on a self-adapted BP algorithm.First,the input and output samples for the network were determined through autoregression analysis,then the hidden units using the trial-and-error method and an ANN model with a structure of 6:10:1 were determined for the simulation and prediction of dynamics of groundwater;finally,a computer program was made with VB according to the improved BP algorithm.The validations of the model show that the precision of the simulation and prediction is high.This model can be applied to the forecast of groundwater dynamics.The predictions after 2002 indicate a continuing decline of groundwater level in the regions,which should be controlled in time.