CAJ | 학술논문

地下水水位动态预测对农田土壤盐渍化防治、地下水地表水资源的合理调度具有十分重要的意义。以新疆和静县某地下水观测井为研究对象，选择月均蒸发量、气温和灌溉量3个因素作为 BP神经网络模型的输入量，利用遗传算法优化神经网络的权值与阈值，建立地下水水位的遗传 BP神经网络预测模型。结果表明：遗传 BP神经网络模型能较好表达地下水位与主控因素之间的非线性关系，预测结果与实测值之间的平均绝对百分比误差为0.0403，测试样本的网络输出值与网络目标值的相关系数达0.9673，模型预测效果较佳。研究结果为区域地下水的开发利用与保护提供参考依据。
지하수수위동태예측대농전토양염지화방치、지하수지표수자원적합리조도구유십분중요적의의。이신강화정현모지하수관측정위연구대상，선택월균증발량、기온화관개량3개인소작위 BP신경망락모형적수입량，이용유전산법우화신경망락적권치여역치，건입지하수수위적유전 BP신경망락예측모형。결과표명：유전 BP신경망락모형능교호표체지하수위여주공인소지간적비선성관계，예측결과여실측치지간적평균절대백분비오차위0.0403，측시양본적망락수출치여망락목표치적상관계수체0.9673，모형예측효과교가。연구결과위구역지하수적개발이용여보호제공삼고의거。
The dynamic groundwater level prediction has very important significance for reasonable scheduling of surface water and groundwater resource and farmland soil salinization prevention. The paper selects the observation wells in Jing county of Xinjiang for studying purpose,selects the average monthly evaporation,temperature and irrigation amount of 3 factors as the input of BP neural networK model,using genetic algorithm to optimize the weight and threshold of the neural networK,the ge-netic BP neural networK to establish the prediction model of groundwater level. The results showed that:the genetic BP neural networK model can better express the nonlinear relation between groundwater level and the main control factors,the mean abso-lute percentage error between the predicted results and the measured value is 0. 0403,the test sample networK output value of the correlation coefficient and the networK target value of 0. 9673,the effect of forecast model is better. The results provide a reference for the development and utilization of groundwater.