基于机器学习预测化学-生物混合污泥厌氧发酵过程中的产酸性能

赵可; 李天乐; 刘常杰; 平倩

doi:10.13205/j.hjgc.202606001

基于机器学习预测化学-生物混合污泥厌氧发酵过程中的产酸性能

doi: 10.13205/j.hjgc.202606001

赵可¹,
李天乐¹,
刘常杰¹,
平倩^1,2, ,

1. 同济大学环境科学与工程学院水污染控制与资源绿色循环全国重点实验室, 上海 200092;
2. 上海污染控制与生态安全研究院, 上海 200092

基金项目:

国家自然科学基金项目（C类）“CaO2促进化学-生物混合污泥两相厌氧消化产甲烷及碳代谢机制研究”（52100158）；同济大学第十九期大学生创新实践训练计划（SITP19）“基于机器学习预测化学-生物混合污泥厌氧发酵过程中产酸性能的研究”

详细信息

作者简介:
赵可(2005-),女,本科生。2353460@tongji.edu.cn

通讯作者:
平倩(1991-),女,助理教授,主要研究方向为有机固废处理与资源化。pingqian@tongji.edu.cn

计量
- 文章访问数: 17
- HTML全文浏览量: 1
- PDF下载量: 4
- 被引次数: 0
出版历程
- 收稿日期: 2025-10-23
- 录用日期: 2025-11-06
- 修回日期: 2025-10-30
- 网络出版日期: 2026-07-06

Machine learning-based prediction of acidogenic performance in anaerobic fermentation of chemical-biological sewage sludge

ZHAO Ke¹,
LI Tianle¹,
LIU Changjie¹,
PING Qian^{1,2
, ,}

1. State Key Laboratory of Water Pollution Control and Green Resource Recycling, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, China;
2. Shanghai Institute of Pollution Control and Ecological Security, Shanghai 200092, China

摘要

摘要: 我国市政污泥厌氧资源化效率普遍低于欧美发达国家水平，化学除磷工艺的广泛应用使得剩余污泥中铁盐、铝盐等化学沉淀含量增加，形成化学-生物混合污泥，降低了厌氧发酵产酸效率。为解析影响该类污泥产酸性能的主要因素，研究构建了高精度预测模型以支撑资源化利用。通过整合文献与实验数据，获取多种厌氧发酵条件下的产酸效能指标，系统对比了BP神经网络、自适应模糊神经网络、支持向量机、K近邻算法和随机森林5种机器学习模型的预测性能，并基于最优模型进行特征重要性解析。结果表明：随机森林模型预测效果最优，测试集决定系数高达0.9463，显著优于其他模型，且过拟合风险最低，针对处理厌氧发酵此类高维、非线性、多因素耦合问题具有强大优势。基于该模型的特征重要性分析揭示，pH值与挥发性悬浮固体（VSS）是驱动产酸效能的关键变量，铝盐的影响程度高于铁盐，因此，工程应用中可按照“调pH、稳有机质、控铝盐”的路径进行工艺优化。研究结果不仅为化学-生物混合污泥的高效资源化提供了智能预测工具，更明确了具体的工艺优化方向，对推动污泥处理技术的精准化与智能化发展具有重要意义。
- 市政污泥 /
- 短链脂肪酸 /
- 模型预测 /
- 厌氧发酵 /
- 资源循环
Abstract: The efficiency of recovering resources from municipal sludge through anaerobic processes is generally lower in China than in developed countries. The widespread adoption of chemical phosphorus removal processes has increased the content of chemical precipitates （e.g.， iron and aluminum salts） in waste activated sludge， resulting in the formation of chemical-biological sewage sludge that reduces acidogenic efficiency during anaerobic fermentation. This study aims to identify the key factors affecting the acidogenic performance in such sludge and to develop a high-precision prediction model to facilitate its resource recovery. By integrating literature and experimental data， acidogenic performance indicators under various anaerobic fermentation conditions were obtained. The predictive capabilities of five machine learning models—Backpropagation Neural Network， Adaptive Neuro-Fuzzy Inference System， Support Vector Machine， K-Nearest Neighbors， and Random Forest —were systematically compared， and feature importance was analyzed using the optimal model. The results indicated that the Random Forest model achieved the best predictive performance， with a coefficient of determination as high as 0.9463 on the test set， significantly outperforming the other models while exhibiting minimal overfitting risk. This demonstrated its strong capability in handling the high-dimensional， nonlinear， and multi-factor coupled problems like anaerobic fermentation. Feature importance analysis based on this model revealed that pH and Volatile Suspended Solids （VSS） were the primary drivers of acidogenic efficiency， with the effect of aluminum salts being greater than that of iron salts. Consequently， for engineering applications， process optimization should follow the pathway of "adjusting pH， stabilizing organic matter content， and controlling aluminum salts". This study not only provides an intelligent predictive tool for the efficient resource recovery of chemical-biological hybrid sludge but also clarifies specific directions for process optimization， holding significant importance for promoting the precision and intelligent development of sludge treatment technology.
- municipal sludge /
- short-chain fatty acids /
- model prediction /
- anaerobic fermentation /
- resource recovery

HTML全文

参考文献(37)

[1]	CHEN R J，DONG B，DAI X H. Overview and prospect on application of hydrothermal treatment on sewage sludge harmlessness[J]. Environmental Engineering，2023，41（9）：201-209. 陈仁杰，董滨，戴晓虎. 水热技术在污泥无害化处理中的应用及展望[J]. 环境工程，2023，41（9）：201-209.
[2]	CHEN S S，YANG D H，PANG W H，et al. Main influencing factors and mechanisms of anaerobic transformation of excess sludge in China[J]. Chemical Industry and Engineering Progress，2020，39（4）：1511-1520. 陈思思，杨殿海，庞维海，等. 我国剩余污泥厌氧转化的主要影响因素及影响机制研究进展[J]. 化工进展，2020，39（4）：1511-1520.
[3]	KIM J O，CHUNG J. Inhibitory effects of inorganic and organic coagulants on acidogenic fermentation[J]. KSCE Journal of Civil Engineering，2015，19（3）：572-577.
[4]	XU Y，LU Y，ZHENG L，et al. Perspective on enhancing the anaerobic digestion of waste activated sludge[J]. Journal of Hazardous Materials，2020，389：121847.
[5]	CHEN Y，QI W，MO R，et al. Synergistic enhancement of anaerobic digestion and resource recovery via extraction of gelatinous extracellular polymeric substances from waste activated sludge：performance and underlying mechanism[J]. ACS Sustainable Chemistry& Engineering，2025，13（28）：11153-11164.
[6]	ZHANG Z，PING Q，GAO D，et al. Effects of ferric-phosphate forms on phosphorus release and the performance of anaerobic fermentation of waste activated sludge[J]. Bioresource Technology，2021，323：124622.
[7]	MOHAMED Y，ELGHADBAN A，LEI H I，et al. Quantum machine learning regression optimisation for full-scale sewage sludge anaerobic digestion[J]. npj Clean Water，2025，8（1）：17.
[8]	GUPTA R，ZHANG L，HOU J，et al. Review of explainable machine learning for anaerobic digestion[J]. Bioresource Technology，2023，369：128468.
[9]	MEOLA A，WEINRICH S. Full-scale dynamic anaerobic digestion process simulation with machine and deep learning algorithms at intra-day resolution[J]. Applied Energy，2025，390：125781.
[10]	ALEJO L，ATKINSON J，GUZMÁN-FIERRO V，et al. Effluent composition prediction of a two-stage anaerobic digestion process：machine learning and stoichiometry techniques[J]. Environmental Science and Pollution Research，2018，25（21）：21149-21163.
[11]	GANESHAN P，BOSE A，LEE J，et al. Machine learning for high solid anaerobic digestion：performance prediction and optimization[J]. Bioresource Technology，2024，400：130665.
[12]	WANG L，LONG F，LIAO W，et al. Prediction of anaerobic digestion performance and identification of critical operational parameters using machine learning algorithms[J]. Bioresource Technology，2020，298：122495.
[13]	LONG F，WANG L，CAI W，et al. Predicting the performance of anaerobic digestion using machine learning algorithms and genomic data[J]. Water Research，2021，199：117182.
[14]	PARK J G，JUN H B，HEO T Y. Retraining prior state performances of anaerobic digestion improves prediction accuracy of methane yield in various machine learning models[J]. Applied Energy，2021，298：117250.
[15]	XU W，LONG F，ZHAO H，et al. Performance prediction of ZVI-based anaerobic digestion reactor using machine learning algorithms[J]. Waste Management，2021，121：59-66.
[16]	LI J，ZHANG L，LI C，et al. Data-driven based in-depth interpretation and inverse design of anaerobic digestion for CH₄-rich biogas production[J]. ACS ES&T Engineering，2022，2（4）：642-652.
[17]	DE CLERCQ D，WEN Z，FEI F，et al. Interpretable machine learning for predicting biomethane production in industrial-scale anaerobic co-digestion[J]. Science of the Total Environment，2020，712：134574.
[18]	KAZEMI P，STEYER J P，BENGOA C，et al. Robust data-driven soft sensors for online monitoring of volatile fatty acids in anaerobic digestion processes[J]. Processes，2020，8（1）：67.
[19]	YU S I，JEONG H，SHIN J，et al. Simulation models of microbial community，pH，and volatile fatty acids of anaerobic digestion developed by machine learning[J]. Journal of Water Process Engineering，2024，60：105225.
[20]	ZOU J，ZHANG L，WANG L，et al. Enhancing phosphorus release from waste activated sludge containing ferric or aluminum phosphates by EDTA addition during anaerobic fermentation process[J]. Chemosphere，2017，171：601-608.
[21]	YAN W，CHEN Y，SHEN N，et al. The influence of a stepwise pH increase on volatile fatty acids production and phosphorus release during Al-waste activated sludge fermentation[J]. Bioresource Technology，2021，320：124276.
[22]	PING Q，ZHANG Z，DAI X，et al. Novel CaO₂ beads used in the anaerobic fermentation of iron-rich sludge for simultaneous short-chain fatty acids and phosphorus recovery under ambient conditions[J]. Bioresource Technology，2021，322：124553.
[23]	ZHANG C，ZHANG Z，ZHANG S，et al. Methionine-based insights into C-S-Fe-P transformations in anaerobic co-digestion of sludge containing iron-phosphorus compounds[J]. Water Research，2025，287：124458.
[24]	ZHANG Z，PING Q，GUO W，et al. A novel approach using protein-rich biomass as co-fermentation substrates to enhance phosphorus recovery from FePs-bearing sludge[J]. Water Research，2022，218：118479.
[25]	ZHANG Z，LIU R，ZHENG W，et al. Specialized genera and niche partitioning promote the biosynthesis of short-chain fatty acids in anaerobic cofermentation of sewage sludge and protein-rich waste[J]. Environmental Research，2025，271：121034.
[26]	PING Q，ZHANG J，TANG R，et al. Effect of surfactants on phosphorus release and acidogenic fermentation of waste activated sludge containing different aluminium phosphate forms[J]. Chemosphere，2022，287：132213.
[27]	ZHANG Z，LIU R，ZHENG W，et al. Enhancing phosphorus release and acidogenesis via endogenous iron-activated peroxydisulfate and corn gluten meal in sludge anaerobic fermentation[J]. Chemical Engineering Journal，2025，512：162583.
[28]	LU K，PING Q，LU Q，et al. Understanding roles of humic substance and protein on iron phosphate transformation during anaerobic fermentation of waste activated sludge[J]. Bioresource Technology，2022，355：127242.
[29]	PING Q，ZHENG M，DAI X，et al. Metagenomic characterization of the enhanced performance of anaerobic fermentation of waste activated sludge with CaO₂ addition at ambient temperature：fatty acid biosynthesis metabolic pathway and CAZymes[J]. Water Research，2020，170：115309.
[30]	PING Q，LU X，LI Y，et al. Effect of complexing agents on phosphorus release from chemical-enhanced phosphorus removal sludge during anaerobic fermentation[J]. Bioresource Technology，2020，301：122745.
[31]	CAI D，LI D，CHEN X，et al. Innovative application of CaO₂ in two-phase anaerobic digestion to enhance methane production from waste activated sludge：condition optimization and mechanistic insights[J]. Chemical Engineering Journal，2025，504：158911.
[32]	WANG Y，HUNTINGTON T，SCOWN C D. Tree-based automated machine learning to predict biogas production for anaerobic co-digestion of organic waste[J]. ACS Sustainable Chemistry& Engineering，2021，9（38）：12990-13000.
[33]	CHOI S，KIM S I，YULISA A，et al. Proactive prediction of total volatile fatty acids concentration in multiple full-scale food waste anaerobic digestion systems using substrate characteristics with machine learning and feature analysis[J]. Waste and Biomass Valorization，2023，14（2）：593-608.
[34]	AVINASH L S，MISHRA A. Comparative evaluation of artificial intelligence based models and kinetic studies in the prediction of biogas from anaerobic digestion of MSW[J]. Fuel，2024，367：131545.
[35]	WANG R，ZHA B. A research on the optimal design of BP neural network based on improved GEP[J]. International Journal of Pattern Recognition and Artificial Intelligence，2019，33（3）：1959007.
[36]	YANGIN GOMEC C，SPEECE R E. The role of pH in the organic material solubilization of domestic sludge in anaerobic digestion[J]. Water Science and Technology，2003，48（3）：143-150.
[37]	CHENG Y，SHI Z，SHI Y，et al. Biochar promoted microbial iron reduction in competition with methanogenesis in anaerobic digestion[J]. Bioresource Technology，2023，387：129561.