基于回归分析和神经网络模型的地下水位预测

Groundwater Level Prediction based on Regression Analysis and Neural Network Model

为了指导和管理地下水资源的合理开发利用,通过建立回归模型和神经网络模型,结合定襄县实际情况,采用1993—2016年的系列资料,对定襄县地下水位进行了预测和分析,详细讨论了2种模型的确定和模型参数的选取过程,比较了2种模型的特点和适用条件。研究表明,前期影响因素分析必不可少,监测资料的可靠性对2种模型的精度有重要影响,资料系列年限足够长将使回归模型系数更准确,使神经网络模型得到全局最优结果,2种模型预测结果可以相互对比和印证。研究成果对定襄县地下水资源的合理开发利用具有指导意义,对其他区域地下水位预测具有参考价值。

In order to guide and manage the rational development and utilization of groundwater resources,through the establishment of regression models and neural network models,combined with the actual situation of Dingxiang County,using a series of data from 1993 to 2016,the groundwater level in Dingxiang County is predicted and analyzed,the determination of the two models and the selection process of model parameters are discussed in detail,and the characteristics and applicable conditions of the two models are compared.The research shows that the analysis of early influencing factors is essential,and the reliability of monitoring data has an important impact on the accuracy of the two models.If the data series is long enough,the coefficients of the regression model will be more accurate,and the neural network model will get the global optimal results.The prediction results of the two models can be compared and confirmed with each other.The research results have guiding significance for the rational development and utilization of groundwater resources in Dingxiang County,and have reference value for the prediction of groundwater level in other cegions.