This research presents a machine learning-driven optimization framework designed to determine optimal purging strategies for polymer electrolyte membrane fuel cells (PEMFCs) operating with dead-ended anode (DEA) configurations. While DEA operation pro...
This research presents a machine learning-driven optimization framework designed to determine optimal purging strategies for polymer electrolyte membrane fuel cells (PEMFCs) operating with dead-ended anode (DEA) configurations. While DEA operation provides significant benefits by eliminating hydrogen recirculation systems and achieving superior hydrogen utilization rates, it encounters challenges related to voltage reduction caused by accumulated water and inert gas permeation in the anode.
To tackle these challenges, a one-dimensional electrochemical dynamic model was constructed to capture the characteristics of voltage decline, water buildup, and electrochemical surface area (ECSA) reduction under diverse operational and purging scenarios. Comprehensive simulation datasets were generated spanning multiple operating parameters, including current density, temperature, and pressure conditions. Using these datasets, a Gaussian Process Regression (GPR) surrogate model was developed to estimate voltage drop (ΔV), hydrogen utilization (HU), and energy efficiency (η), demonstrating prediction accuracy with R² exceeding 0.96 and RMSE below 1.1%. This surrogate model was integrated with the NSGA-II multi-objective optimization algorithm to obtain the Pareto front characterizing the trade-off between ΔV and HU, enabling the identification of optimal purging strategies that maximize overall energy efficiency. For typical operating parameters (0.8 A/cm², 70 °C, 0.8 bar), the optimization procedure identified an optimal purging configuration with an interval of 73.3 s and duration of 0.020 s.
The proposed framework functions as a computational design instrument for determining globally optimal purging strategies that balance voltage stability with high hydrogen utilization while maximizing energy efficiency, offering practical applications in establishing purging guidelines, system architecture design, and controller implementation for DEA-PEMFCs