Stack-level engineering and predictive diagnostics of a modular plant–microbial fuel cell system using artificial neural networks

Abstract

Plant-microbial fuel cells (PMFCs) are an emerging bioelectrochemical technology for sustainable energy generation that harnesses root exudates and soil microbial metabolism. Their operation shares electrochemical characteristics with hydrogen fuel cell stacks, particularly in terms of electrode polarization and voltage balancing. Here, we present the integrated design, fabrication, and predictive modeling of a modular six-cell PMFC stack ( Patent in process ), 3D-printed with a PETG-based housing where each individual cell features an integrated soil chamber. The system is deployed under real-agricultural conditions within the rhizosphere of Psidium guajava L. The soil matrix is characterized by an acidic pH (5.09), high organic matter content (27.7%), and substantial nitrate availability (53.9 mg/kg), which together support a peak open-circuit voltage of 1.159 V and a maximum power density of 4.15 mW m −2 per cell. A deep artificial neural network (20–10) predicts voltage from 34 ANN input variables (R 2 = 0.9866; MAE = 0.0165 V), enabling real-time performance verification through integrated monitoring and optical signaling.