Optimizing serpentine PEM fuel cell performance: AI-enhanced multi-objective analysis

• Three new flow patterns are proposed for a fuel cell with the serpentine channel. • Artificial intelligence and machine learning algorithms were used for modeling. • Multi-objective optimization with a genetic algorithm was employed. • The optimal humidity of 14.57 % was obtained. • The use of proposed models can significantly increase the cell's performance. Proton Exchange Membrane (PEM) fuel cells represent a promising solution for clean energy generation, particularly when utilizing hydrogen. This study introduces and investigates three innovative flow orientations—termed models A, B, and C—within the serpentine-channeled PEM fuel cell architecture, each presenting varying angles between the anode and cathode sides. Employing a genetic algorithm coupled with multi-objective optimization techniques, we aimed to enhance cathode input humidity, subsequently reducing input power consumption while concurrently augmenting output power. The optimization process leveraged deep neural networks and polynomial regression to model and refine data, enabling single-objective Nelder-Mead and multi-objective genetic algorithm (NSGA-II) approaches. Notably, achieving an optimal 14 % humidity significantly bolstered output power and current density, while effectively curbing power consumption and water content within cell layers. Furthermore, this research explores the impact of these novel flow patterns on fundamental parameters. Through comprehensive analysis, models A, B, and C revealed substantial improvements in current density and power production, notably model C showcasing a 0.62 W/cm^2 maximum power density.