摘要
以地下水环境模拟参数随机场中的离子质量浓度为例,以研究BP神经网络方法(BPNN)应用于地下水环境模拟参数随机场空间变异性的可能性。将所有数据分为独立的训练和检验数据集,用没有参与建模的22组数据进行验证,并用最佳BPNN模型进行区域空间分布预测图的绘制。结果表明:①BPNN方法的插值结果与观测值的相关系数达到0.952,平均偏差为1.438,协方差为14.052,取得了较好的模拟效果,且估值效果明显好于普通克里格法。②从最佳模型区域空间分布预测图来看,该方法能比较客观地刻画地下水环境模拟参数随机场中离子质量浓度的空间分布状况。这一实际应用表明,BPNN方法可以较好地描述地下水环境模拟参数随机场的空间分布规律。
The paper focuses on the possibility of applying BP Neural Networks（BPNN） to describe the spatial distribution of the parameters random field in ordinary groundwater environmental simulation.The ion mass concentration of parameters in the random fields for groundwater environment simulation were selected for the study,the samples were divided into training and validation datum sets,22 sampling data were used as validated data in the BPNN model.The results of the model showed： ①The average deviation is 1.4380,the covariance is 14.0524,and correlation coefficient between the observed values and the modeling values is 0.9515.The prediction precision of the BPNN model is better than ordinary kriging.②The regional prediction map of optimal BP model can describe the spatial distribution of the parameters random field in ordinary groundwater environmental simulation.The result shows that the method of BPNN can be used to describe the spatial distribution rule of the parameters random field in ordinary groundwater environmental simulation.
出处
《水资源与水工程学报》
2010年第5期20-24,共5页
Journal of Water Resources and Water Engineering
基金
教育部科学技术研究重点项目(03028)
北京林业大学振兴计划人才培养专项课题(200202013)
