Integrating functionalized catalysts and machine learning to optimize microbial fuel cells for clean energy applications

Abstract

The dual challenge of clean energy generation and wastewater treatment has intensified interest in microbial fuel cells (MFCs) as sustainable, bioelectrochemical systems. In this study, four low-cost cathode catalysts based on polyaniline (PANI) derivatives functionalized with diethanolamine (DEA) and ethylenediamine (EDA), along with phthalocyanine and copper-phthalocyanine (CuPc), were synthesized and evaluated. Compared to the MFC using carbon paper, the CuPc-based system (MFC4) achieved a ∼7.25-fold increase in power density (408.3 mW m⁻²vs. 56.3 mW m⁻²) and the highest coulombic efficiency (∼18%), highlighting enhanced electron transfer and ORR activity. Meanwhile, the PANI–EDA-based MFC (MFC5) demonstrated the highest COD removal efficiency (∼90%), reflecting improved microbial compatibility. These performance improvements were closely related to the surface chemistry of the catalysts. In CuPc, the central Cu–N coordination provides stable active sites that mimic enzymatic centers and facilitate oxygen adsorption and reduction. Functionalization of PANI with EDA and DEA introduces amine and hydroxyl groups that increase surface hydrophilicity, promote proton and electron transport, and create additional catalytic sites, thus significantly boosting ORR kinetics and biofilm–electrode interactions. To complement experimental observations, machine learning models (CatBoost and XGBoost) were applied to explore the relationship between key system variables and power density. Among different machine learnings, XGBoost exhibited the highest predictive accuracy (R² = 0.959 and RMSE = 19.90), which enabled highly accurate predictions for the modeling of the MFC output. This work presents a comparative and surface-engineered approach to optimizing non-precious cathode catalysts, contributing to the development of efficient and integrated MFC technologies.