机器学习模型精细表征填埋场渗漏风险及其不确定性

能昌信; 张慧敏; 孙晓晨; 徐亚; 刘景财; 刘玉强

PDF(2008 KB)

中国环境科学 ›› 2025, Vol. 45 ›› Issue (9) : 4986-4996.

固体废物

机器学习模型精细表征填埋场渗漏风险及其不确定性

能昌信^1,2, 张慧敏^1,2, 孙晓晨¹, 徐亚^2,3, 刘景财², 刘玉强²

作者信息 +

Machine learning models for fine characterization of landfill leakage and uncertainty analysis

NAI Chang-xin^1,2, ZHANG Hui-min^1,2, SUN Xiao-chen¹, XU Ya^2,3, LIU Jing-cai², LIU Yu-qiang²

Author information +

文章历史 +

摘要

针对传统基于解析解与Monte Carlo耦合的填埋场地下水污染风险预测方法难以准确刻画渗漏不确定性的问题,提出了一种高性能仿真模拟与不确定性分析耦合的填埋场渗漏风险及其不确定性表征技术.为解决传统不确定性分析框架下,高仿真模拟数值模型计算负荷大,不确定性分析效率低的问题,引入粒子群优化算法(PSO)与极限梯度提升算法(XGBoost)模型,通过学习精细模拟模型计算获得的多组样本,建立替代模型来预测;同时,为了解决PSO-XGBoost对抽样数量的过度依赖,提出了改进的抽样方法.实验结果表明,与以Landsim为代表的解析解-monte carlo耦合算法相比,本研究构建基于数值模拟模型样本的机器学习替代模型更好的捕捉了参数不确定性对结果的影响,预测的污染浓度分布区间平均为传统模型的1.36倍;基于PSO-XGBoost的替代模型及改进的抽样方法,相比于其他机器学习模型,预测精度(R²=0.99)和稳定性更好,且模型训练阶段计算时间减少70%,大幅提升了不确定性分析的效率.基于该模型的应用研究表明,第10a特征污染物COD超标的概率达到56.43%,存在超标可能的距离为54m,表明研究区存在一定污染风险.研究成果为填埋场渗漏污染风险及不确定性分析提供了高效、精准的科学方法,并为进一步开展相关研究提供了借鉴.

Abstract

This study addresses the limitations of traditional landfill groundwater pollution risk prediction methods by integrating high-performance simulation with uncertainty analysis to more accurately capture the effects of concealed leakage. The limitations of high computational load and low efficiency were addressed by combining a Particle Swarm Optimization (PSO) algorithm with an Extreme Gradient Boosting (XGBoost) model. By learning from multiple samples generated by detailed simulation models, a surrogate model was established to improve prediction efficiency. Additionally, an improved sampling method was proposed to reduce the reliance of PSO-XGBoost on sample size. The results indicate that the machine learning surrogate model developed in this study better captures the impact of parameter uncertainty compared to traditional analytical–Monte Carlo methods such as Landsim. The predicted pollution concentration range is 1.36 times greater than that of conventional models. The PSO-XGBoost surrogate model, combined with the improved sampling method, outperforms other machine learning models in prediction accuracy (R²= 0.99) and stability, while reducing the computational time for model training by 70%, significantly improving the efficiency of uncertainty analysis. The application research based on this model shows that the probability of the characteristic pollutant COD exceeding the standard in the 10th year reaches 56.43%, and the possible distance of exceeding the standard is 54meters, indicating that there is a certain pollution risk in the study area. These findings provide an efficient and precise scientific method for analyzing landfill leakage risks and uncertainties and offer valuable references for further related research.

导出引用

能昌信, 张慧敏, 孙晓晨, 徐亚, 刘景财, 刘玉强. 机器学习模型精细表征填埋场渗漏风险及其不确定性[J]. 中国环境科学. 2025, 45(9): 4986-4996

NAI Chang-xin, ZHANG Hui-min, SUN Xiao-chen, XU Ya, LIU Jing-cai, LIU Yu-qiang. Machine learning models for fine characterization of landfill leakage and uncertainty analysis[J]. China Environmental Science. 2025, 45(9): 4986-4996

中图分类号： X523

参考文献

[1] Xu Y, Xue X, Dong L, et al. Long-term dynamics of leachate production, leakage from hazardous waste landfill sites and the impact on groundwater quality and human health [J]. Waste Management, 2018,82:156-166.
[2] Vinti G, Bauza V, Clasen T, et al. Health risks of solid waste management practices in rural Ghana:A semi-quantitative approach toward a solid waste safety plan [J]. Environmental Research, 2023, 216:114728.
[3] Aydin M E, Aydin S, Bahadir M, et al. Toxicity assessment of landfill leachates with a battery of bioassays [J]. Fresenius Environmental Bulletin, 2015,24(11):3584-3589.
[4] 张帅,周弋铃,彭靖宇,等.生活垃圾填埋场渗滤液渗漏场地污染风险评价体系 [J].环境科学研究, 2024,37(07):1583-1591.DOI:10. 13198/j.issn.1001-6929.2024.03.04. Zhang S, Zhou G L, Peng J Y, et al. Contamination risk assessment system for leachate leakage from municipal solid waste landfill sites [J]. Research of Environmental Sciences, 2024,37(7):1583-1591. DOI:10.13198/j.issn.1001-6929.2024.03.04.
[5] Sun X C, Xu Y, Liu Y Q, et al. Evolution of geomembrane degradation and defects in a landfill:Impacts on long-term leachate leakage and groundwater quality [J]. Journal of Cleaner Production, 2019,224:335-345.
[6] Saheli P T, Rowe R K, Petersen E J, et al. Diffusion of multiwall carbon nanotubes through a high-density polyethylene geomembrane [J]. Geosynthetics International, 2017,24(2):184-197.
[7] 张鲁玉.基于深度学习的危废填埋渗漏风险预测 [D].山东:山东工商学院, 2023.DOI:10.27903/d.cnki.gsdsg.2023.000154. Zhang L Y. Risk Prediction of Hazardous Waste Landfil Leakage Based on Deep Learning [D]. Shandong:Shandong Technology and Business University, 2023.DOI:10.27903/d.cnki.gsdsg.2023.0001:54.
[8] Sheng D, Wen X, Wu J, et al. Comprehensive Probabilistic Health Risk Assessment for Exposure to Arsenic and Cadmium in Groundwater [J]. Environmental Management, 2021,67(4):779-792.
[9] Ciriello V, Lauriola I, Bonvicini S, et al. Impact of Hydrogeological Uncertainty on Estimation of Environmental Risks Posed by Hydrocarbon Transportation Networks [J]. Water Resources Research, 2017,53(11):8686-8697.
[10] Demissie Y, Valocchi A, Cai X, et al. Parameter estimation for groundwater models under uncertain irrigation data [J]. Groundwater, 2015,53(4):614-625.
[11] 莫绍星.基于深度学习的地下水模拟高维不确定性分析和反演 [D].南京:南京大学, 2019.DOI:10.27235/d.cnki.gnjiu.2019.000108. Mo S X. Towards efficient high-dimensional uncertainty quantification and inverse analysis in groundwater modeling using deep learning [D]. Nanjing: Nanjing University, 2019.DOI:10.27235/ d.cnki.gnjiu.2019.000108.
[12] Lahmar S, Maalmi M, Idchabani R. Investigating adaptive sampling strategies for optimal building energy performance using artificial neural networks and kriging surrogate models [J]. Journal of Building Engineering, 2024,82:108341.
[13] 陈骏骏.基于深度学习的地下水反应性溶质运移模型参数反演研究 [D].吉林:吉林大学, 2023.DOI:10.27162/d.cnki.gjlin.2023.000764. Chen J J. Deep learning based inverse modeling of groundwater reactive transport models [D]. Jilin: Jilin University, 2023.DOI:10. 27162/d.cnki.gjlin.2023.000764.
[14] 罗苑萍,孙世泰,黄旭斌,等.基于机器学习的变电站地下水埋深智能预报预警模型 [J/OL].水利水电技术(中英文)[2024-10-30]. Luo Y P, Sun S T, Huang X B, et al. Intelligent forecasting and early warning models for substation drainage systems based on [J/OL]. Water Resources and Hydropower Engineering machine learning, [2024-10-30].
[15] Alsaif A,Abbas Y M. Interpretable constitutive compressive stress- strain model for rubberized aggregate concrete-Integrating comprehensive empirical database and efficient XGBoost ensemble learning [J]. Case Studies in Construction Materials, 2024,21:e03382.
[16] 董广齐,卢文喜,范越,等.不确定性条件下地下水污染监测井网优化设计——基于XGBoost替代模型 [J].中国环境科学, 42(5): 2144-2152. Dong G Q, Lu W X, Fan Y, et al. Optimal design of groundwater pollution monitoring network under uncertainty [J]. China Environmental Science. 42(5):2144-2152.
[17] 罗建男,李雪利,王鹤,等.基于局部-全局混合自适应替代模型的地下水污染源反演识别 [J].中国环境科学, 2023,43(7):3664-3671. Luo J N, Li X L, Wang H, et al. Groundwater pollution sources inversion based on local-global hybrid adaptive surrogate model [J]. China Environmental Science. 2023,43(7):3664-3671.
[18] 葛渊博,卢文喜,白玉堃,等.基于SSA-BP与SSA的地下水污染源反演识别 [J].中国环境科学, 2022,42(11):5179-5187. Ge Y B, Lu W X, Bai Y K, et al. Inversion and identification of groundwater pollution sources based on SSA-BP and SSA [J]. China Environmental Science. 2022,42(11):5179-5187.
[19] 贾顺卿,卢文喜,李久辉,等.基于U-D分解卡尔曼滤波地下水污染源溯源辨识 [J].中国环境科学, 2021,41(2):713-719. Jia S Q, Lu W X, Li J H, et al. Inversion identification of groundwater contamination source based on U-D factorization Kalman filter [J]. China Environmental Science. 2021,41(2):713-719.
[20] Lin G,Tartakovsky A M.An efficient, high-order probabilistic collocation method on sparse grids for three-dimensional flow and solute transport in randomly heterogeneous porous media [J]. Advances in Water Resources, 2009,32(5):712-722.
[21] Kavvas M, Chen Y, Strelkoff T. Performance assessment of a subsurface drip irrigation system using the HELP model [J]. Irrigation Science, 2002,21(3):129-137.
[22] Nai C, Tang M, Liu Y, et al. Potentially contamination and health risk to shallow groundwater caused by closed industrial solid waste landfills: Site reclamation evaluation strategies [J]. Journal of cleaner production, 2021,286:125402.
[23] Wu C M, Yeh T C J, Zhu J F, et al. Traditional analysis of comparing apples to oranges [J]. Water Resources Research, 2005,41(9)W09402.
[24] Berger K U. On the current state of the Hydrologic Evaluation of Landfill Performance (HELP) model [J]. Waste management, 2015,38: 201-209.
[25] 王新港,杨昱,徐祥健,等.单井抽出-回渗同步循环地下水水力控制模型研究 [J].环境科学研究, 2023,36(01):180-187.DOI:10. 13198/j.issn.1001-6929.2022.08.05. Wang X G, Yang Y, Xu X J, et al. Hydraulic control model research on single well pumping-recharge synchronous cyclical groundwater [J]. Research of Environmental Sciences, 2023,36(1):180-187.DOI:10. 13198/j.issn.1001-6929.2022.08.05.
[26] Pan Y, Zeng X, Xu H, et al. Use of stacked Gaussian processes regression method to improve prediction of groundwater solute transport model [J]. Journal of Hydrology, 2023,620:129530.
[27] 王静.基于深度学习的地下水问题数值模拟方法 [D].武汉:武汉大学, 2021. Wang J. Numerical simulation of ground water problem using a deep learning approach [D]. Wuhan:Wuhan University, 2021.
[28] 张雨佳,杨蕴,龚绪龙,等.油水混合层CO₂地质封存与利用数值模拟及参数全局敏感性分析 [J/OL].中国环境科学, 1-12[2025-07-25]. https://doi.org/10.19674/j.cnki.issn1000-6923.20250321.003. Zhang Y J, Yang Y, Gong X L, et al. Numerical simulation and parameter global sensitivity analysis of CO₂ geological storage and utilization in oil-water mixed layers [J]. China Environmental Science. 1-12[2025-07-25].https://doi.org/10.19674/j.cnki.issn1000-6923.20250321.003.
[29] Dai X M,Liu L X,Cheng Z.Elucidating price variability drivers in highway electromechanical equipment using CV predictions with PSO-XGBoost[J].Alexandria Engineering Journal,2024,109:754-767.
[30] 邵萌,潘正中,孙金伟,等.基于EMD-PSO-BP模型的短期潮流流速预测 [J].中国海洋大学学报(自然科学版), 2024,54(11):134- 141.DOI:10.16441/j.cnki.hdxb.20220321.
[31] Shao M, Pan Z Z, Sun J W, et al. Short-term tidal current speed prediction based on EMD-PSO-BP model [J]. Periodical of Ocean University of China, 2024,54(11):134-141.
[32] 邓田丰,李梓铭,张少波,等.基于机器学习和时间序列算法融合的北京地区花粉浓度预测技术研究 [J/OL].中国环境科学,1-23[2025- 07-25].https://doi.org/10.19674/j.cnki.issn1000-6923.20250624.001. Deng T F, Li Z M, Zhang S B, et al. Research on pollen concentration prediction technology in Beijing based on the fusion of machine learning and time series algorithms. [J]. China Environmental Science. 1-23[2025-07-25].https://doi.org/10.19674/j.cnki.issn1000-6923.20250624.001.
[33] Zhang J, Cao X, Li C, et al. Experimental analysis of combustion characteristics of corn starch dust clouds under the action of unilateral obstacles and machine learning modeling based on PSO-XGBoost [J]. Advanced Powder Technology, 2024,35(11): 104641.
[34] Sheikhi S, Kostakos P. Safeguarding cyberspace: Enhancing malicious website detection with PSOoptimized XGBoost and firefly-based feature selection [J]. Computers & Security, 2024,142:103885.
[35] Bo Y, Guo X, Liu Q, et al. Prediction of tunnel deformation using PSO variant integrated with XGBoost and its TBM jamming application [J]. Tunnelling and Underground Space Technology, 2024,150:105842.
[36] Luo J, Ma X, Ji Y, et al. Review of machine learning-based surrogate models of groundwater contaminant modeling [J]. Environmental Research, 2023:117268.
[37] 王永财,万华伟,高吉喜,等.基于深度学习语义分割的数码相机图像草地植被盖度估算 [J/OL].环境科学研究[2024-10-30].DOI: 10.13198/j.issn.1001-6929.2024.09.02. Wang Y C, Wan H W, Gai J X, et al. Grassland vegetation coverage estimation in digital camera images based on deep learning semantic segmentation [J/OL]. Research of Environmental Sciences. [2024- 10-30].DOI:10.13198/j.issn.1001-6929.2024.09.02.
[38] 徐亚宁,卢文喜,王梓博,等.考虑参数和边界条件不确定性的地下水污染随机模拟 [J].中国环境科学, 2022,42(7):3244-3253. Xu Y N, Lu X W, Wang Z B, et al. Stochastic simulation of groundwater pollution considering uncertainty of parameters and boundary conditions [J]. China Environmental Science. 2022,42(7): 3244-3253.
[39] 刘志刚,郭婧婷,佘祺锐,等.北京安定生活垃圾填埋场扩容边坡稳定性分析 [C] //中国环境科学学会环境工程分会.中国环境科学学会2021年科学技术年会——环境工程技术创新与应用分会场论文集(二).北京:《工业建筑》杂志社有限公司, 2021:234-238,105.DOI:10. 26914/c.cnkihy.2021.022070. Liu Z G, Guo J T, Yu Q R, et al. Slope stability analysis of anding municipal solid waste landfill in Beijing [C]. Environmental Engineering Branch of the Chinese Society for Environmental Sciences. Proceedings of the sub - session on innovation and application of environmental engineering technology at the 2021 scientific and technological annual conference of the Chinese Society for Environmental Sciences (II). Beijing: Industrial Construction Magazine Co., Ltd., 2021:234-238,105.DOI:10.26914/c.cnkihy.2021. 022070.
[40] 张永贺;刘亚淘;蓝智钢.北京市安定卫生填埋场渗沥液处理工艺 [J].环境卫生工程, 2013,21(6):61-64. Zhang Y H, Liu Y T, Lan Z G, et al. Leachate treatment technology of anding waste sanitary landfill site in Beijing [J]. Environmental Sanitation Engineering. 2013,21(6):61-64.
[41] 张弦.基于LSTM-BP的填埋场污染扩散监测方法研究 [D].烟台:山东工商学院, 2022.DOI:10.27903/d.cnki.gsdsg.2022.000091. Zhang X. Research on monitoring method of landfill pollution diffusion based on LSTM-BP [D]. Yantai: Shandong Technology and Business University, 2022.DOI:10.27903/d.cnki.gsdsg.2022. 000091.
[42] 陈亚松,刘家雯,赵云鹏,等.基于机器学习的人工湿地出水水质预测与影响因素 [J].中国环境科学, 2025,45(06):3161-3170.DOI: 10.19674/j.cnki.issn1000-6923.2025.0085. Chen Y S, Liu J W, Zhao Y P, et al. Prediction of effluent water quality and analysis of influencing factors in constructed wetlands based on machine learning. [J]. China Environmental Science. 2025,45(6): 3161-3170.DOI:10.19674/j.cnki.issn1000-6923.2025.0085.