灰色神经网络在地下水动态预测中的应用

以周至201号井为例，选取降雨量、蒸发量、单位面积的引灌水量及人工开采量4个地下水位的主要影响因素为预报因子，地下水位作为输出样本，建立BP神经网络模型。以2002-2011年4个序列的数据分别建立新陈代谢GM(1,1)模型，得到2012-2014年的预测值。再将各新陈代谢GM(1,1)模型得到的4个预报因子的预测值作为BP神经网络的输入，得到的输出即为最终2012-2014年地下水位的预测值。结果表明，灰色理论和BP神经网络耦合模型具有较高的预测精度，可为地下水的动态预报提供参考。

A Dynamic Prediction of Groundwater Based on Coupled Model of GM(1,1) Model and BP Model

For well 201 in Zhouzhi, taking rainfall capacity, evaporation capacity, irrigation capacity amount per unit of area and artificial exploitation quantity as input samples and taking groundwater level as output sample to establish BP Model. Using 4 sequences of data from 2002 to 2011 establishes GM(1,1) model and gains predictive value from 2012 to 2014. Taking the predictive value as input factors in BP Model, and the final output is the predictive value from 2012 to 2014 of groundwater. The results show that coupled model of GM(1,1) model and BP Model has high accuracy and can provide references for groundwater level regime forecast.