人工神经网络在我国西南地区岩溶地下水污染评价中的应用  

Application of artificial neural network to karst groundwater pollution assessment in Southwest China

作　　者：刘伟 刘昊洋 朱天龙 谢浩 邹胜章[2,3] 杨国丽 马立山[1] 李军[1] LIU Wei;LIU Haoyang;ZHU Tianlong;XIE Hao;ZOU Shengzhang;YANG Guoli;MA Lishan;LI Jun(Hebei University of Architecture,Zhangjiakou,Hebei 075000;Key Laboratory of Karst Dynamics,Ministry of Natural Resources/Guangxi,Institute of Karst Ceology,Chinese Academy of Geological Sciences,Guilin,Guangxi 541004;International Research Center on Karst Under the Auspices of United Nations Educational,Scientific and Cultural Organization,Guilin,Guangxi 541004)

机构地区：[1]河北建筑工程学院,河北张家口075000 [2]中国地质科学院岩溶地质研究所,自然资源部/广西岩溶动力学重点实验室,广西桂林541004 [3]联合国教科文组织国际岩溶研究中心,广西桂林541004

出　　处：《河北建筑工程学院学报》2024年第3期127-135,共9页Journal of Hebei Institute of Architecture and Civil Engineering

基　　金：河北省教育厅青年基金项目“张家口典型水环境污染综合评价体系集成与应用研究”(2022QNJS05);河北省教育厅青年基金项目“基于生态理念的校园雨水及中水综合利用研究”(QN2020424)。

摘　　要：岩溶地下水是人类生活的重要用水来源,预测岩溶地下水污染水平对可持续开发利用地下水资源和保护岩溶地区生态环境具有重要意义。为获取更可靠的岩溶地下水污染水平预测结果,利用我国西南部典型岩溶区105组地下水样品,在内梅罗综合污染指数评价量化结果基础上,分别采用BP(Back Propagation)神经网络、SVM(Support Vector Machine)和T-S(Takagi-Sugeno)模糊神经网络进行地下水污染等级预测,比较不同人工神经网络污染评价预测结果的可靠性。利用90组岩溶地下水样品进行人工神经网络训练,剩余15组岩溶地下水样品进行污染等级预测比拟。结果表明,BP神经网络模型预测结果与内梅罗综合污染指数评价结果相似性为80%,SVM相似性为73.3%,T-S模糊神经网络相似性为66.7%。BP神经网络预测的岩溶地下水污染等级结果与内梅罗综合污染指数评价量化结果最为相近,且优于SVM和T-S模糊神经网络预测结果。基于此,在我国西南部岩溶地下水污染评价研究中,建议优先选用BP人工神经网络进行岩溶地下水污染水平预测。Karst groundwater is an important source of water for human life.Predicting the pollution level of karst groundwater is of great significance for the sustainable development and utilization of groundwater resources and the protection of the ecological environment in karst areas.In order to obtain more representative evaluation results of karst groundwater pollution,a total of 105 groundwater samples were collected from typical karst areas,southwestern China.On the basis of the quantitative results of Nemero Comprehensive Pollution Index,the BP(Back Propagation)neural network,SVM(Support Vector Machine)and T-S(Takagi-Sugeno)fuzzy neural network were used to predict groundwater pollution level.Further,the reliabilities of three artificial neural network conducted in pollution evaluation and prediction were compared.In these processes,90karst groundwater samples were applied to train artificial neural network,and remaining 15karst groundwater samples were employed for predicting the pollution level.The results exhibited that the similarity between the prediction results of BP neural network model and the evaluation results of Nemero comprehensive pollution index is 80%,the similarity of SVM is 73.3%,and the similarity of T-S fuzzy neural network is 66.7%.The prediction result of karst groundwater pollution level from BP neural network is better than that from the SVM,and also better than that from the T-S.Moreover,the obtained prediction result from BP neural network is similar to the result from the Nemero Comprehensive Pollution Index.Based on this,in the study of karst groundwater pollution evaluation in southwest China,it is suggested that BP artificial neural network should be preferred to predict the pollution level of karst groundwater.

关 键 词：岩溶地下水 污染评价 内梅罗污染指数 人工神经网络 

分 类 号：X523[环境科学与工程—环境工程]