Abstract
Proton exchange membrane fuel cell (PEMFC) performance is governed by tightly coupled electrochemical reactions and multi-phase transport processes occurring within the membrane electrode assembly (MEA). In particular, the thicknesses of the proton exchange membrane (PEM), catalyst layer (CL), and gas diffusion layer (GDL) jointly control protonic resistance, catalyst utilization, oxygen accessibility, and liquid water evacuation. However, most available studies still optimize MEA layers independently, despite the fact that realistic PEMFC operation is dominated by nonlinear coupling between charge transport and two-phase water dynamics. In this work, a high-fidelity three-dimensional two-phase PEMFC model was developed in OpenFOAM using the openFuelCell2 framework to resolve multi-species transport, electrochemical kinetics, conjugate heat transfer, and multiphase water transport across all MEA layers. A systematic multi-parameter dataset was generated by simultaneously varying PEM thickness (25–200 μm), CL thickness (3–20 μm), and GDL thickness (150–350 μm) under fixed operating conditions. To enable efficient exploration of the nonlinear high-dimensional design space, a surrogate-assisted artificial neural network–genetic algorithm (ANN–GA) optimization framework was constructed. The ANN surrogate was trained using CFD-generated polarization data and subsequently coupled with a real-coded GA to maximize power density. The integrated optimization identified an optimal MEA configuration of tPEM = 25 μm, tCL = 14.0037 μm, and tGDL = 167.6356 μm, achieving a maximum predicted power density of 0.3785 W/cm2 at 0.8733 A/cm2. High-fidelity CFD verification yielded 0.3748 W/cm2, corresponding to a deviation below 1.01%, confirming the reliability of the surrogate-driven optimization. Mechanistic analysis reveals that the optimum arises from the combined minimization of protonic ohmic losses (thin PEM), enhancement of electrochemically active reaction volume without severe oxygen diffusion limitation (moderate CL), and reduction of cathode-side mass transport resistance while suppressing excessive liquid water accumulation (thin-to-moderate GDL). The proposed CFD–ANN–GA framework provides a computationally scalable pathway for high-dimensional MEA structural optimization and offers quantitative design guidance for next-generation PEMFC development under two-phase operating conditions.
| Original language | English |
|---|---|
| Article number | 101912 |
| Journal | Energy Conversion and Management: X |
| Volume | 30 |
| DOIs | |
| Publication status | Published - May 2026 |
UN SDGs
This output contributes to the following UN Sustainable Development Goals (SDGs)
-
SDG 7 Affordable and Clean Energy
Keywords
- Artificial neural network
- Genetic algorithm
- OpenFOAM
- Proton exchange membrane fuel cell
- Surrogate modelling
- Two-phase CFD
Fingerprint
Dive into the research topics of 'Coupled thickness optimization of PEM, CL, and GDL in Two-Phase PEMFCs using OpenFOAM-Based 3D CFD and Sensitivity-Guided ANN–GA surrogate modeling'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver