Supervised Learning Algorithm Reveals the Mechanistic Role of Extracellular Polymeric Substances in Biofouling of an Anaerobic Electrochemical Membrane Bioreactor

Extracellular polymeric substances (EPS)-related biofouling in anaerobic electrochemical membrane bioreactors (AnEMBR) is commonly mitigated, yet the underlying mechanisms require further elucidation. This study developed an electrochemical anaerobic membrane bioreactor to mitigate membrane fouling by modulating the EPS properties. Compared to the control AnMBR, the AnEMBR achieved a 67.7% reduction in fouling at a voltage of 1.2 V. Confocal laser scanning microscopy (CLSM) further revealed a 47.5% decrease in the fluorescence intensity of EPS-protein foulants within the fouling layer. Furthermore, tightly bound EPS (TB-EPS) concentrations in the AnEMBR decreased by 34.2%, contributing to the control of irreversible membrane fouling. To investigate the intrinsic role of EPS in fouling, a predictive membrane fouling model was developed using a supervised learning algorithm trained on experimental EPS data sets. After hyperparameter optimization, the model demonstrated excellent predictive performance with an R2 of 0.92. Shapley Additive exPlanations analysis identified TB-EPS proteins as the most critical factor influencing membrane fouling. Partial dependence plots further elucidated the marginal effects of different EPS fractions on membrane fouling and their respective partial dependence characteristics. Finally, quartz crystal microbalance with dissipation monitoring (QCM-D) validated the model’s predictions, revealing that TB-EPS in the AnEMBR formed a less viscoelastic and more removable biofouling layer compared to the AnMBR. This study highlights the mechanistic role of EPS in AnEMBR biofouling and underscores their significance for advancing sustainable strategies to mitigate membrane fouling in wastewater treatment systems.