Physics-Guided Modeling and Artificial Intelligence for Proton Exchange Membrane Fuel Cells in Extreme Environments

Abstract

Proton exchange membrane fuel cells (PEMFCs) deployed in aviation, underwater, and space applications encounter different combinations of environmental and mission constraints. These include altitude-dependent oxygen supply, flight-load transients, closed reactant loops, constrained heat rejection, and gravity-dependent two-phase transport. The resulting changes in reactant supply, water–thermal balance, and degradation complicate system integration. Intelligent methods can support rapid modeling, state estimation, health management, and constrained control, yet their applicability depends on the operating environment and validation evidence. This review connects environment-specific mechanisms with modeling and deployment requirements, distinguishing three evidence levels: direct studies under representative extreme conditions, studies reproducing individual relevant stressors, and transferable methods established under conventional PEMFC conditions. The synthesis identifies gaps in data representativeness, cross-domain generalization, physical consistency, uncertainty assessment, and closed-loop validation. An integrated physics–data–decision framework is then proposed to connect environment-aware modeling, state and health estimation, risk-aware control, and staged validation. Its purpose is to guide method selection and identify the evidence needed for reliable onboard operation in aerial, underwater, and space systems.