Review on Data Science Models Used in Diagnosis of Polymer Electrolyte Membrane Fuel Cells

ABSTRACT

Proton exchange membrane fuel cells (PEMFCs) are important in the move towards clean energy and decarbonization of automotive power systems around the world. Adoption faces challenges such as changing component performance and degradation in real‐world conditions. Data science models and artificial intelligence (AI) methods, including machine learning, neural networks, and hybrid physics‐informed approaches, are applied to improve the reliability and efficiency of PEMFCs. Neural networks help with fault detection, spatial mapping of current distribution, and prediction of performance decline. Learning reinforcement is used to optimize cell humidity and identify thermal hotspots in single cell and fuel‐cell stacks, which show the need for standardized and benchmarked AI‐based diagnostic protocols. This review examines and compares current data science models for diagnosis in different operating scenarios, evaluates their success in forecasting chemical, electrochemical, and thermal failures, and explores the challenges and performance of these predictive tools in automotive fuel‐cell applications. The findings support the development of regulatory standards and help speed up the industrial adoption of AI‐driven and data science–based diagnostic solutions for fuel‐cell technologies.