基于配线法的BP神经网络求解Theis模型

Computation of Theis Model with Curve Fitting Method-based BP Artificial Neural Network

摘要:

利用非稳定流抽水试验资料确定含水层水文地质参数的配线法在具体应用中存在较大的随意性,用一种改进的反向传播(BP)神经网络方法来进行承压完整井非稳定地下水运动Theis公式中的水文地质参数识别,在一定程度上解决了现有的BP神经网络方法求解含水层参数中存在训练区间选择、网络拓扑结构复杂、网络泛化性能较低和过度拟合等问题.实例计算结果表明,提出的基于配线法的BP神经网络是求解水文地质参数的一种高效方法,对其他水文地质问题求解具有推广应用价值.

Abstract:

Curve fitting method is the most common method to determine the parameters of the transient flow pumping tests,but its main drawback is that curves are often rely solely on individual manipulation.Based on curve fitting method,a modified BP artificial neural network is used to solve the approximately analytical solutions of Theis equation for the problem of the unsteady flow of fully penetrating wells in confined aquifers,and the curve fitting-based BP artificial neural network is proposed.It avoids inappropriate trained range.It also reduces the network topology with a simpler artificial neural network structure,as this can reduce training time,improve network generalization and prevent over-fitting.The case calculating results indicate that the proposed BP artificial neural network is an efficient method to solve hydrogeological parameters by utilizing the pumping test.Furthermore,the proposed BP artificial neural network could be used widely in the solution for hydrogeological issues.