基于BP神经网络的地下水动态预测

Prediction of Groundwater Dynamics Based on the BP Neural Network

人工神经网络是一种高度非线性的并行分布处理系统,采用人工神经网络预测宝鸡市的地下水位动态变化趋势,取1995—2007年研究区内的降水入渗补给量、河道渗漏补给量、人工开采量和闸坝蓄水渗漏量作为输入因子,建立BP模型,用于模拟2008的年地下水位埋深,并与传统的灰色模型进行比较,结果表明:BP神经网络的相对误差介于0.07%～1.98%,相对于灰色模型(0.13%～6.41%)具有较高的预测精度,可为该灌区地下水位的动态预报提供参考。

Artificial neural network is a highly nonlinear parallel distributed processing system. It was used to predict groundwater level in Baoji City. Precipitation infiltration quantity, river leakage recharge, groundwa- ter withdrawal and dam water leakage from 1995 to 2007 were adopted as the input factors to establish BP model to simulate water table depth in 2008. It showed that BP neural network model had high accuracy and the relative error was between 0.07% and 1.98% in comparison with Grey Model （relative error was from 0.13% to 6.41%）. Overall, BP neural network model can provide references for groundwater level regime forecast in irrigation zone of Baoji.