机器学习辅助的双膜耦合氨氮回收过程优化
作者： 游伟明， 唐函， 黄飞云， 姚婧梅， 韩乐*
单位： 重庆大学  环境与生态学院, 重庆  400045
关键词： 氨氮回收； 机器学习； 膜； 耦合
DOI号： 10.16159/j.cnki.issn1007-8924.2026.03.016
分类号： TQ028；TP181
出版年,卷(期):页码： 2026, 46(3):162-170

摘要：

氨氮的高效回收对构建可持续氮循环至关重要。道南渗析-渗透蒸馏（DD-OD）双膜耦合系统以浓差自驱低耗回收氨氮，但复杂的传质过程与显著的非线性特征使其难以建模以优化调控工艺系统。本研究引入机器学习（ML）方法，构建了基于CatBoost的氨氮去除率与回收率预测模型。经优化后，两模型在测试集上的拟合优度（R2）分别达0.97和0.95。通过沙普利解释（SHAP）与偏依赖图（PDP）分析，揭示了运行时间、进料浓度等关键参数对系统的非线性影响规律。进一步构建贝叶斯优化框架，实现了多目标协同优化。在进料室NH+4-N浓度52 mmol/L、运行时间28.0 h及驱动离子与目标离子浓度比（ρ值）1.92的最优工况下，去除率与回收率分别达92.28%与88.03%。研究表明，ML方法可有效应对复杂环境过程的非线性耦合特性，为新型环保系统的智能运行提供支撑。

Efficient recovery of ammonia-nitrogen is essential for establishing a sustainable nitrogen cycle. The Donnan dialysis-osmotic distillation (DD-OD) dual-membrane coupling system recovers ammonia-nitrogen via concentration-driven and low-energy mechanisms, but its complex mass transfer processes and significant non-linear characteristics render it difficult to model for optimizing and regulating the process system. This study introduced machine learning (ML) methods to develop a CatBoost-based prediction model for ammonia-nitrogen removal and recovery rates. Following optimization, the model achieved goodness-of-fit (R2) values of 0.97 and 0.95 on the test set, respectively. Through shapley additive explanations (SHAP) and partial dependence plot (PDP) analyses, the non-linear influence of key parameters, such as operating time and feed concentration, was elucidated. Furthermore, a Bayesian optimization framework was constructed to achieve multi-objective collaborative optimization. Under the optimal conditions-a feed chamber NH+4-N concentration of 52 mmol/L, an operating time of 28.0 h, and a driving-to-target ion concentration ratio (ρ) of 1.92-the removal and recovery rates reached 92.28% and 88.03%, respectively. This research demonstrates that ML methods can effectively address the non-linear coupling characteristics of complex environmental processes, providing a robust foundation for the intelligent operation of novel environmental protection systems. 

基金项目：

重庆市科技创新与应用发展专项重点项目（CSTB2024TIAD-KPX0084）； 新重庆青年创新人才项目（2024NSCQ-QNCXX0233）

作者简介：
第一作者简介： 游伟明（2001-），男，福建福清人，硕士研究生，研究方向为膜法水处理、机器学习辅助的膜智能设计等.*通讯作者，E-mail：lehan@cqu.edu.cn

参考文献：

［1］Mingolla S, Rosa L. Low-carbon ammonia production is essential for resilient and sustainable agriculture［J］. Nat Food, 2025, 6(6): 610-621.
［2］Farghali M, Chen Z, Osman A I, et al. Strategies for ammonia recovery from wastewater: A review［J］. Environ Chem Lett, 2024, 22(6): 2699-2751.
［3］Cruz H, Law Y Y, Guest J S, et al. Mainstream ammonium recovery to advance sustainable urban wastewater management［J］. Environ Sci Technol, 2019, 53(19): 11066-11079.
［4］Beckinghausen A, Odlare M, Thorin E, et al. From removal to recovery: An evaluation of nitrogen recovery techniques from wastewater［J］. Appl Energy, 2020, 263: 114616.
［5］Yan T, Ye Y, Ma H, et al. A critical review on membrane hybrid system for nutrient recovery from wastewater［J］. Chem Eng J, 2018, 348: 143-156.
［6］Chen C, Han M, Yao J, et al. Donnan dialysis-osmotic distillation (DD-OD) hybrid process for selective ammonium recovery driven by waste alkali［J］. Environ Sci Technol, 2021, 55(10): 7015-7024.
［7］Luo T, Abdu S, Wessling M. Selectivity of ion exchange membranes: A review［J］. J Membr Sci, 2018, 555: 429-454.
［8］Dutta A, Kalam S, Lee J. Elucidating the inherent fouling tolerance of membrane contactors for ammonia recovery from wastewater［J］. J Membr Sci, 2022, 645: 120197.
［9］Yang K, Qin M. Understanding ammonia and water transport in direct contact membrane distillation toward selective ammonia recovery［J］. ACS EST Eng, 2024, 4(6): 1321-1330.
［10］Rall D, Schweidtmann A M, Kruse M, et al. Multi-scale membrane process optimization with high-fidelity ion transport models through machine learning［J］. J Membr Sci, 2020, 608: 118208.
［11］Tayyebi A, Alshami A S, Tayyebi E, et al. Machine learning -driven surface grafting of thin-film composite reverse osmosis (TFC-RO) membrane［J］. Desalination, 2024, 579: 117502.
［12］Jeong N, Chung T H, Tong T. Predicting micropollutant removal by reverse osmosis and nanofiltration membranes: Is machine learning viable ［J］. Environ Sci Technol, 2021, 55(16): 11348-11359.
［13］Ma J, Xu H, Wang A, et al. Machine learning-guided underlying decisive factors of high-performance membrane distillation system: Membrane properties, operation conditions and solution composition［J］. Sep Purif Technol, 2023, 327: 124964.
［14］Gao H, Zhong S, Zhang W, et al. Revolutionizing membrane design using machine learning-bayesian optimization［J］. Environ Sci Technol, 2022, 56(4): 2572-2581.
［15］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(6):3161-3170.
［16］Buuren S V, Groothuis-Oudshoorn K. Mice: Multivariate imputation by chained equations in R［J］. Int J Biostat, 2014, 45(2):1-67.
［17］Sun J O, Hua T W, Guan Y F, et al. Predicting and understanding the performance of polyamide nanofiltration membrane for Li/Mg selective separation based on machine learning［J］. Water Res, 2025, 285: 124140.
［18］Li Z, Yuan X, Zhang C, et al. Long-term degradation prediction method for proton exchange membrane fuel cells based on hybrid transfer learning［J］. Int J Hydrog Energy, 2025, 106: 781-789.
［19］Wang M, Shi G M, Zhao D, et al. Machine learning-assisted design of thin-film composite membranes for solvent recovery［J］. Environ Sci Technol, 2023, 57(42): 15914-15924.
［20］Qu D, Sun D, Wang H, et al. Experimental study of ammonia removal from water by modified direct contact membrane distillation［J］. Desalination, 2013, 326: 135-140.
［21］Rosentreter H, Scope C, Oddoy T, et al. Monovalent selective ion exchange membranes: A review on preparation processes, applications, performance criteria and sustainability aspects［J］. Desalination, 2025, 599: 118412.
［22］Dai Z, Chen C, Li Y, et al. Hybrid donnan dialysis-electrodialysis for efficient ammonia recovery from anaerobic digester effluent［J］. Environ Sci Ecotechnol, 2023, 15: 100255.
［23］Rodrigues M, Sleutels T, Kuntke P, et al. Exploiting Donnan dialysis to enhance ammonia recovery in an electrochemical system［J］. Chem Eng J, 2020, 395: 125143.
［24］Narayen D, Van Berlo E, Van Lier J B, et al. Recovery of sulfuric acid and ammonia from scrubber effluents using bipolar membrane electrodialysis: Effect of pH and temperature［J］. Sep Purif Technol, 2024, 338: 126605.
［25］Chen C, Dong T, Han M, et al. Ammonium recovery from wastewater by donnan dialysis: A feasibility study［J］. J Clean Prod, 2020, 265: 121838.
［26］Jung O, Saravia F, Wagner M, et al. Quantifying concentration polarization -Raman microspectroscopy for in-situ measurement in a flat sheet cross-flow nanofiltration membrane unit［J］. Sci Rep, 2019, 9(1): 15885.
［27］Kywe P P, Ratanatamskul C. Influences of permeate solution and feed pH on enhancement of ammonia recovery from wastewater by negatively charged PTFE membranes in direct contact membrane distillation operation［J］. ACS Omega, 2022, 7(31): 27722-27733.
［28］Zhu Z, Dong S, Zhang H, et al. Bayesian optimization-enhanced reinforcement learning for self-adaptive and multi-objective control of wastewater treatment［J］. Bioresour Technol, 2025, 421: 132210.

服务与反馈：
【Download】【加入收藏】

《膜科学与技术》编辑部 地址：北京市朝阳区北三环东路19号蓝星大厦 邮政编码：100029 电话：010-80492417/010-80485372 传真：010-80485372邮箱：mkxyjs@163.com

京公网安备11011302000819号