【摘要】 依据水均衡原理,导出承压水漏斗区任意一点水位与其影响因素之间的复杂的非线性关系,在此基础上提出承压水漏斗水位时空分布预报的BP神经网络模型。该模型具有分布参数模型的特征,且不需用到区域的水文地质参数。最后针对某实例进行了模型设计及预报分析,通过对部分观测井后期实测数据的训练,优选出双隐层的网络结构及其网络参数。随后用这些观测井的前3年数据进行了检验,并对其他观测井数据进行预报。计算表明,该模型对地下水位拟合与预报的合格率较高,可以获得研究区域某一时刻水位在空间上的分布。更多还原

【Abstract】 According to the water balance principle a complicated nonlinear relationships between groundwater stage and relevant influencing factors at arbitrary position of confined water cones region is deduced, and a BP network model for forecasting the temporal and spatial distribution of groundwater stage is suggested. The model possesses the characteristics of distributed parameter model, but it does not require hydrogeological parameters. The MatLab tools box provides the BP network with computation modules for various training functions. An application example is given, in which the network with double hidden layer structure and relevant parameters are optimized by training according to the data obtained from some observation wells, and the validity of the model is verified by the data in latest three years. Thus, the spatial distribution of stage at any moment can be obtained. The comparison of forecasting stages with observation data obtained from other wells indicates that the proposed model has sufficient high qualified ratio.更多还原