基于可解释性机器学习的芬顿工艺降解有机污染物速率的研究

于林堂; 陈咚咚; 陶翠翠; 朱腾义

PDF(1262 KB)

中国环境科学 ›› 2025, Vol. 45 ›› Issue (8) : 4294-4302.

水污染与控制

基于可解释性机器学习的芬顿工艺降解有机污染物速率的研究

于林堂, 陈咚咚, 陶翠翠, 朱腾义

作者信息 +

Research on the degradation rate of organic pollutants in fenton process based on explainable machine learning

YU Lin-tang, CHEN Dong-dong, TAO Cui-cui, ZHU Teng-yi

Author information +

文章历史 +

摘要

采用机器学习模型,包括多元线性回归(MLR)和轻量级梯度提升机(LGB)算法,预测芬顿工艺中57种有机污染物的降解效率.通过SHAP方法对模型进行机理解释,识别了影响降解效率的关键因素.研究结果表明,LGB模型在预测精度上(R²_adj = 0.969, Q²_LOO= 0.925, R²_ext= 0.844)优于MLR模型(R²_adj=0.831, Q²_LOO=0.802, R²_ext=0.861).SHAP分析揭示了温度、分子三维结构和原子电离能力是影响降解效率的主要因素.本研究为优化芬顿工艺的操作条件和提升降解效率提供了科学依据,对水处理领域的研究和实践具有重要的指导意义.

Abstract

Machine learning models, including multiple linear regression (MLR) and decision tree machine learning (LGB) algorithms, were used to predict the degradation efficiency of 57 organic pollutants in the Fenton process. Mechanistic interpretation of the models was performed by SHAP method to identify the key factors affecting the degradation efficiency. The results showed that the LGB model (R²_adj = 0.969, Q²_LOO = 0.925, R²_ext = 0.844) outperformed the MLR model (R²_adj = 0.831, Q²_LOO = 0.802, R²_ext = 0.861) in terms of prediction accuracy. The SHAP analysis revealed that temperature, molecular three-dimensional structure, and atomic ionization capacity were the main factors affecting the degradation efficiency. This study provides a scientific basis for optimizing the operating conditions of the Fenton process and enhancing the degradation efficiency, which is of great significance in guiding the research and practice in the field of water treatment.

导出引用

于林堂, 陈咚咚, 陶翠翠, 朱腾义. 基于可解释性机器学习的芬顿工艺降解有机污染物速率的研究[J]. 中国环境科学. 2025, 45(8): 4294-4302

YU Lin-tang, CHEN Dong-dong, TAO Cui-cui, ZHU Teng-yi. Research on the degradation rate of organic pollutants in fenton process based on explainable machine learning[J]. China Environmental Science. 2025, 45(8): 4294-4302

中图分类号： X131

参考文献

[1] Alvarez P J J, Chan C K, Elimelech M, et al. Emerging opportunities for nanotechnology to enhance water security [J]. Nature Nanotechnology, 2018,13(8):634-641.
[2] Eliasson J. The rising pressure of global water shortages [J]. Nature, 2015,517(7532):6.
[3] Schwarzenbach R P, Escher B I, Fenner K, et al. The challenge of micropollutants in aquatic systems [J]. Science, 2006,313(5790): 1072-1077.
[4] Bello M M, Raman A A A, Asghar A. A review on approaches for addressing the limitations of Fenton oxidation for recalcitrant wastewater treatment [J]. Process Safety and Environmental Protection, 2019,126:119-140.
[5] 谢谨裕,王卓雯,高伟杰,等.钴基气体扩散电极强化电芬顿处理磺胺噻唑钠[J].中国环境科学, 2023,43(1):122-131. Xie J Y,Wang Z W,Gao W J, et al. Cobalt-based gas diffusion electrodes enhance the E-Fenton for sulfathiazole sodium treatment [J]. China Environmental Science, 2023,43(1):122-131.
[6] 杜茂华,李皓芯,任婧,等.改性阴极生物电芬顿系统降解罗丹明B [J].中国环境科学, 2021,41(4):1681-1688. Du M H, Li H X, Reng J, et al. Electricity production and rhodation B degradation in bioelectric fenton system with modified cathode [J]. China Environmental Science, 2021,41(4):1681-1688.
[7] 辛丽红,晏彩霞,聂明华,等.微量铜离子联合碳酸氢盐类芬顿体系降解水中双酚A [J].中国环境科学, 2023,43(3):1186-1196. Xin L H, Yan C X, Nie M H, et al. Degradation of bisphenol A by Fenton-like system using trace copper ions combined with bicarbonate in water [J]. China Environmental Science, 2023,43(3):1186-1196.
[8] Cheng M, Zeng G M, Huang D L, et al. Hydroxyl radicals based advanced oxidation processes (AOPs) for remediation of soils contaminated with organic compounds: A review [J]. Chemical Engineering Journal, 2016,284:582-598.
[9] Hou X J, Huang X P, Jia F L, et al. Hydroxylamine promoted goethite surface fenton degradation of organic pollutants [J]. Environmental Science & Technology, 2017,51(9):5118-5126.
[10] Jubb A M, Gierczak T, Baasandorj M, et al. Methyl- perfluoroheptene-ethers (CH₃OC₇F₁₃): Measured OH radical reaction rate coefficients for several isomers and enantiomers and their atmospheric lifetimes and global warming potentials [J]. Environmental Science & Technology, 2014,48(9):4954-4962.
[11] 高歌,张文晴,魏歆倪,等.新型优先管控污染物筛选研究进展[J].中国环境科学, 2024,44(11):6472-6483. Gao G, Zhang W Q, Wei X N, et al. Research progress on screening of emerging priority controlled contaminants [J]. China Environmental Science, 2024,44(11):6472-6483.
[12] Louridas P, Ebert C. Machine Learning [J]. Ieee Software, 2016,33(5): 110-115.
[13] Fedik N, Zubatyuk R, Kulichenko M, et al. Extending machine learning beyond interatomic potentials for predicting molecular properties (vol 6, pg 653, 2022) [J]. Nature Reviews Chemistry, 2022, 6(12):916-916.
[14] 何家乐,陈景文,王杰琼,等.环境水体中化学品光化学持久性的预测模型[J].科学通报, 2024,69(6):731-745. He J L, Chen J W, Wang J Q, et al. Prediction models for photochemical persistence of chemicals in aquatic environments (in Chinese). Chin Sci Bull, 2024,69:731-745.
[15] Qin L T, Zhang X, Chen Y H, et al. Predictive QSAR models for the toxicity of disinfection byproducts [J]. Molecules, 2017,22(10):11.
[16] Cheng Z W, Yang B W, Chen Q C, et al. Quantitative relationships between molecular parameters and reaction rate of organic chemicals in Fenton process in temperature range of 15.8°C-60°C [J]. Chemical Engineering Journal, 2018,350:534-540.
[17] Ghanbari-Adivi E. A new machine learning model for predicting the water quality index [J]. Modeling Earth Systems and Environment, 2024,10(4):5635-5667.
[18] Cheng Z W, Yang B W, Chen Q C, et al. 2D-QSAR and 3D-QSAR simulations for the reaction rate constants of organic compounds in ozone-hydrogen peroxide oxidation [J]. Chemosphere, 2018,212:828- 836.
[19] Jia L J, Shen Z M, Guo W M, et al. QSAR models for oxidative degradation of organic pollutants in the Fenton process [J]. Journal of the Taiwan Institute of Chemical Engineers, 2015,46:140-147.
[20] Cheng Z W, Yang B W, Chen Q C, et al. Characteristics and difference of oxidation and coagulation mechanisms for the removal of organic compounds by quantum parameter analysis [J]. Chemical Engineering Journal, 2018,332:351-360.
[21] Fizir M, Richa A, Touil S, et al. Comparing adsorption properties of algerian kaolinite towards phosphate with those of activated carbon: adsorption experiments, molecular modeling and an initiative toward hydroponic wastewaters remediation [J]. Water Air and Soil Pollution, 2024,235(6):19.
[22] Elmorsy M R, Abdel-Latif E, Badawy S A, et al. Molecular geometry, synthesis and photovoltaic performance studies over 2-cyanoacetanilides as sensitizers and effective co-sensitizers for DSSCs loaded with HD-2[J]. Journal of Photochemistry and Photobiology a-Chemistry, 2020,389:8.
[23] Zhu H C, Guo W M, Shen Z M, et al. QSAR models for degradation of organic pollutants in ozonation process under acidic condition [J]. Chemosphere, 2015,119:65-71.
[24] Mauri A. alvaDesc: A tool to calculate and analyze molecular descriptors and fingerprints [M] //Roy K. Ecotoxicological QSARs. New York, NY: Springer US. 2020:801-820.
[25] Bieroza M, Baker A, Bridgeman J. New data mining and calibration approaches to the assessment of water treatment efficiency [J]. Advances in Engineering Software, 2012,44(1):126-135.
[26] Sperandei S. Understanding logistic regression analysis [J]. Biochemia Medica, 2014,24(1):12-18.
[27] Lyu J, Zheng P J, Qi Y, et al. LightGBM-LncLoc: A LightGBM- based computational predictor for recognizing long non-coding RNA subcellular localization [J]. Mathematics, 2023,11(3):13.
[28] 魏佳妹,袁书娟,孔闪闪,等.轻梯度提升机算法的发展与应用[J].计算机工程与应用, 2025,61(5):32-42. Wei J M, Yuan S J, Kong S S, et al. Development and application of light gradient boosting machine [J]. Computer Engineering and Applications, 2025,61(5):32-42.
[29] Gupta V, Kumar E. H³O-LGBM: hybrid Harris hawk optimization based light gradient boosting machine model for real-time trading [J]. Artificial Intelligence Review, 2023,56(8):8697-8720.
[30] Zhang D Y, Gong Y C. The comparison of LightGBM and XGBoost coupling factor analysis and prediagnosis of acute liver failure [J]. Ieee Access, 2020,8:220990-221003.
[31] Qin H, Chen J W, Wang Y, et al. Development and assessment of quantitative structure-activity relationship models for bioconcentration factors of organic pollutants [J]. Chinese Science Bulletin, 2009,54(4): 628-634.
[32] Hanser T, Barber C, Guesne S, et al. Applicability domain: Towards a more formal framework to express the applicability of a model and the confidence in individual predictions [M]. Springer International Publishing Ag, Gewerbestrasse 11, Cham, Ch-6330, Switzerland, 2019.
[33] Khoei T T, Slimane H O, Kaabouch N. Deep learning: systematic review, models, challenges, and research directions [J]. Neural Computing & Applications, 2023,35(31):23103-23124.
[34] Rodríguez-Pérez R, Bajorath J. Interpretation of compound activity predictions from complex machine learning models using local approximations and shapley values [J]. Journal of Medicinal Chemistry, 2020,63(16):8761-8777.
[35] Baptista M L, Goebel K, Henriques E M P. Relation between prognostics predictor evaluation metrics andlocal interpretability SHAP values [J]. Artificial Intelligence, 2022,306:22.
[36] Khosravi A, Koury R N N, Machado L, et al. Prediction of hourly solar radiation in Abu Musa Island using machine learning algorithms [J]. J. Clean Prod, 2018,176:63-75.
[37] Chirico N, Gramatica P. Real external predictivity of QSAR Models. Part 2. New intercomparable thresholds for different validation criteria and the need for scatter plot inspection [J]. Journal of Chemical Information and Modeling, 2012,52(8):2044-2058.
[38] Kiralj R, Ferreira M M C. Basic Validation procedures for regression models in QSAR and QSPR studies: Theory and application [J]. Journal of the Brazilian Chemical Society, 2009,20(4):770-787.
[39] 王中钰,陈景文,傅志强,等.QSAR模型应用域的表征方法[J].科学通报, 2022,67(3):255-266. Wang Z Y, Chen J W, Fu Z Q, et al. Characterization of applicability domains for QSAR models [J]. Chinese Science Bulletin, 2022,67(3): 255-266.
[40] 赵清波,徐勇勇,夏结来.logistic回归中高杠杆点的检测[J].中国卫生统计, 1997,(2):19-22. Zhao Q B, Xu Y Y, Xia L J. Detection of high leverage points in logistic regression [J]. Chinese Journal of Health Statistics, 1997,(2): 19-22.
[41] Li Q, Yang H, Hao N, et al. Biodegradability analysis of dioxins through in silico methods: Model construction and mechanism analysis [J]. Journal of Environmental Management, 2023,345:9.
[42] Singh P, Kumar R, Sharma B K, et al. Topological descriptors in modeling malonyl coenzyme A decarboxylase inhibitory activity: N-Alkyl-N-(1,1,1,3,3,3-hexafluoro-2-hydroxypropylphenyl)amide derivatives [J]. Journal of Enzyme Inhibition and Medicinal Chemistry, 2009,24(1):77-85.
[43] Abdulrahman H L, Uzairu A, Uba S. Computer modeling of some anti-breast cancer compounds [J]. Structural Chemistry, 2021,32(2): 679-687.
[44] Mamy L, Patureau D, Barriuso E, et al. Prediction of the fate of organic compounds in the environment from their molecular properties: A review [J]. Critical Reviews in Environmental Science and Technology, 2015,45(12):1277-1377.
[45] Che Z P, Zhang S Y, Shao Y H, et al. Synthesis and quantitative structure-activity relationship (QSAR) study of novel n- arylsulfonyl-3-acylindole arylcarbonyl hydrazone derivatives as nematicidal agents [J]. Journal of Agricultural and Food Chemistry, 2013,61(24):5696-5705.
[46] Prajapati L M, Parmar V K, Patel M J, et al. Development of QSAR model for prediction of fold selectivity of phenyl benzoxazole as estrogen receptor inhibitors [J]. Medicinal Chemistry Research, 2012, 21(11):3681-3686.
[47] Klein C T, Kaiser D, Ecker G. Topological distance based 3D descriptors for use in QSAR and diversity analysis [J]. Journal of Chemical Information and Computer Sciences, 2004,44(1):200-209.
[48] Devinyak O, Havrylyuk D, Lesyk R. 3D-MoRSE descriptors explained [J]. Journal of Molecular Graphics & Modelling, 2014,54: 194-203.