基于ELM模型的浅层地下水位埋深时空分布预测
发布时间：2019-03-03 10:41
【摘要】：选用石家庄平原区补排因子的多种组合为输入参数,利用28眼水井的实测资料作为预测目标值,首次建立基于极限学习机(Extreme learning machine,ELM)的地下水位埋深时空分布预测模型,讨论补排因子在不同缺失情况下对模型精度的影响;利用Arc GIS分析误差空间分布趋势,并与常用的三隐层BP神经网络模型进行对比。结果表明:基于水均衡理论的ELM地下水位埋深模拟模型能够准确反映人类和自然双重影响下地下水系统的非线性关系,模型输入因子中缺失降水量或开采量的模拟结果均方根误差(RMSE)比缺失其余因子的RMSE高2.00倍及以上,同时模型有效系数(E_(ns))和决定系数(R~2)进一步降低;与BP模型相比,ELM模型可使RMSE减小43.6%,误差区间降低46.4%,Ens和R2提高至0.99,且RMSE在空间相同区域上均明显呈现出ELM模型小于BP模型;ELM模型在南部高误差区的移植精度(RMSE低于1.82 m/a,E_(ns)高于0.95)高于BP模型(RMSE超过3.00 m/a,Ens低于0.85);因此,影响地下水位埋深的主导因素是降水量和开采量,且ELM模型在精度、稳定性和空间均匀性上较优,移植预测效果较好,可利用已知资料推求区域空间内其余未知水井的浅层地下水位埋深;该模型可作为水文地质参数及补排资料缺乏条件下浅层地下水位埋深预测的推荐模型。
[Abstract]:The spatial-temporal distribution prediction model of groundwater table depth based on limit learning machine (Extreme learning machine,ELM) is established for the first time by using the measured data of 28 wells as the prediction target value and using various combinations of complementary drainage factors in Shijiazhuang Plain as input parameters. The influence of complementary removal factor on the accuracy of the model is discussed in different cases. The spatial distribution trend of error is analyzed by Arc GIS and compared with the three hidden layer BP neural network model. The results show that the ELM groundwater table depth simulation model based on the water equilibrium theory can accurately reflect the nonlinear relationship between human and natural groundwater systems. The root mean square error (RMSE) of the model input factor is 2.00 times higher than that of the other factors, and the effective coefficient (E _ (ns) and decision coefficient) of the model is further reduced. The root mean square error (RMS) of the model input factor is 2.00 times higher than that of the other factors. Compared with BP model, ELM model reduced RMSE by 43.6%, error interval decreased by 46.4%, Ens and R2 increased to 0.999, and RMSE showed that ELM model was smaller than BP model in the same area of space. The transplant accuracy of the ELM model in the southern high error area (RMSE < 1.82m / a, E _ (ns) > 0.95) was higher than that of the BP model (RMSE > 3.00m / a, Ens < 0.85). Therefore, the main factors affecting the depth of groundwater table are precipitation and mining amount, and the ELM model is better in precision, stability and spatial uniformity, and the effect of transplant prediction is better. The shallow groundwater table depth of the remaining unknown wells in the regional space can be calculated by using the known data. This model can be used as the recommended model for predicting the depth of shallow groundwater table under the condition of lack of hydrogeological parameters and supplementary drainage data.
【作者单位】： 昆明理工大学现代农业工程学院;长沙理工大学水利工程学院;中国农业科学院农业环境与可持续发展研究所;作物高效用水与抗灾减损国家工程实验室;宁乡县水利水电勘测设计院;
【分类号】：P641
本文编号：2433622
[Abstract]:The spatial-temporal distribution prediction model of groundwater table depth based on limit learning machine (Extreme learning machine,ELM) is established for the first time by using the measured data of 28 wells as the prediction target value and using various combinations of complementary drainage factors in Shijiazhuang Plain as input parameters. The influence of complementary removal factor on the accuracy of the model is discussed in different cases. The spatial distribution trend of error is analyzed by Arc GIS and compared with the three hidden layer BP neural network model. The results show that the ELM groundwater table depth simulation model based on the water equilibrium theory can accurately reflect the nonlinear relationship between human and natural groundwater systems. The root mean square error (RMSE) of the model input factor is 2.00 times higher than that of the other factors, and the effective coefficient (E _ (ns) and decision coefficient) of the model is further reduced. The root mean square error (RMS) of the model input factor is 2.00 times higher than that of the other factors. Compared with BP model, ELM model reduced RMSE by 43.6%, error interval decreased by 46.4%, Ens and R2 increased to 0.999, and RMSE showed that ELM model was smaller than BP model in the same area of space. The transplant accuracy of the ELM model in the southern high error area (RMSE < 1.82m / a, E _ (ns) > 0.95) was higher than that of the BP model (RMSE > 3.00m / a, Ens < 0.85). Therefore, the main factors affecting the depth of groundwater table are precipitation and mining amount, and the ELM model is better in precision, stability and spatial uniformity, and the effect of transplant prediction is better. The shallow groundwater table depth of the remaining unknown wells in the regional space can be calculated by using the known data. This model can be used as the recommended model for predicting the depth of shallow groundwater table under the condition of lack of hydrogeological parameters and supplementary drainage data.
【作者单位】： 昆明理工大学现代农业工程学院;长沙理工大学水利工程学院;中国农业科学院农业环境与可持续发展研究所;作物高效用水与抗灾减损国家工程实验室;宁乡县水利水电勘测设计院;
【分类号】：P641
【相似文献】
相关博士学位论文 前2条
1 陈志云;基于RS/GIS的西辽河平原地下水位埋深变化与植被生态的响应研究[D];吉林大学;2012年
2 刘圣;马海盆地荒漠绿洲区生态地下水位对人类活动响应的研究[D];中国地质大学（北京）;2014年
相关硕士学位论文 前1条
1 王琦;保定市市区浅层地下水位埋深预测方法的研究[D];中国地质大学(北京);2016年
，本文编号：2433622
本文链接：https://www.wllwen.com/kejilunwen/diqiudizhi/2433622.html
