基于改进集成学习的矿井地下水污染风险预测研究  

Research on Risk Prediction of Mine Groundwater Pollution Based on Improved Ensemble Learning

作　　者：李婷[1] 李艳军[1] 吕英英 杨娟娟[1] 白岩立 Li Ting;Li Yanjun;Lv Yingying;Yang Juanjuan;Bai Yanli(Yulin Vocational And Technical College,Yulin 719000,China;Yulin Municipal Ecology and Environment Bureau,Yulin 719000,China;Yulin Energy Group Hengshan Coal Power Co.,Ltd.,Yulin 719200,China)

机构地区：[1]榆林职业技术学院,陕西榆林719000 [2]榆林市生态环境局,陕西榆林719000 [3]榆林能源集团横山煤电有限公司,陕西榆林719200

出　　处：《环境科学与管理》2024年第2期178-182,共5页Environmental Science and Management

基　　金：陕西省科协青年人才托举计划项目(20220403)。

摘　　要：在进行矿井地下水污染风险预测过程中,由于选择的特征与污染风险相关性较低,导致预测精度较差,对此,提出基于改进集成学习的矿井地下水污染风险预测研究,首先利用主成分分析法对矿井地下水污染数据特征进行提取,然后利用SOM网络进行矿井地下水数据聚类处理,最后采用ENN模型进行矿井地下水污染风险预测。实验结果表明,所提方法的污染物浓度预测RMSE和MAPE分别为22 mg·L^(-1)与9.26%,矿井地下水污染风险指数与实际值拟合度高,且R 2值较大,说明所提方法的风险预测能力较好,具有实用性。In the process of predicting the risk of groundwater pollution in mines,the low correlation between the selected features and pollution risk leads to poor prediction accuracy.Therefore,a research on predicting the risk of groundwater pollution in mines based on improved ensemble learning is proposed.Firstly,principal component analysis is used to extract the characteristics of groundwater pollution data in mines,and then SOM network is used for clustering processing of groundwater data in mines,Finally,the ENN model is used to predict the risk of groundwater pollution in mines.The experimental results show that the RMSE and MAPE for pollutant concentration prediction of the proposed method are 22 mg·L^(-1)and 9.26%,respectively.The risk index of mine groundwater pollution has a high fitting degree with the actual value,and the R 2 value is relatively large,indicating that the proposed method has good risk prediction ability and practicality.

关 键 词：改进集成学习 污染风险预测 SOM网络 ELMAN神经网络 

分 类 号：X820.4[环境科学与工程—环境工程]