Predicting the performance of microbial electrochemical systems equipped with three-dimensional anodes using machine learning

Three-dimensional (3D) structured materials have emerged as promising anode architectures in microbial electrochemical systems (MESs) due to their ability to enhance biofilm formation and electron transfer. However, accurately predicting MES power output based on the physical and electrochemical properties of 3D anodes remains a major challenge, as conventional experimental and theoretical models are often constrained by limited scalability and poor handling of nonlinear variable interactions. In this study, we present the first application of machine learning (ML) to systematically investigate and predict the power output of MESs equipped with 3D anodes. A comprehensive dataset was compiled from the literature, encompassing all available experimental data on MESs utilizing 3D anode configurations. Multiple ML algorithms were trained and evaluated, among which the XGBoost model exhibited the highest predictive accuracy. Feature importance analysis revealed that anode charge transfer resistance (Rct), ohmic resistance (Ro), and thickness are the three influential variables governing system performance. The effectiveness of the model in predicting power output was further confirmed by actual 3D printing experimental verification. These findings provide insights into the complex relationships between anode structure, electrochemical impedance, and power generation in MESs. The proposed ML-based framework offers a robust and scalable approach to guide the rational design of next-generation 3D anodes for energy-efficient wastewater treatment and resource recovery applications.