Artificial intelligence and machine learning models for electrolyzers and fuel cells optimization: A review on empowering sustainable hydrogen economy

Abstract

The migration to a hydrogen economy necessitates advanced optimization of electrochemical devices such as electrolyzers (ELs) and fuel cells (FCs). This review outlines the change in thinking from traditional, computationally intensive physics-based modeling to artificial intelligence (AI) and machine learning (ML) frameworks. It details how AI/ML architectures facilitate real-time performance prediction via surrogate models, prognostic health management (PHM) for automated fault diagnosis, and adaptive control strategies for operational efficiency. Furthermore, it analyzes emerging hybrid strategies, namely physics-informed neural networks (PINNs) which incorporate governing physical laws to enhance model robustness and data efficiency. Although significant progress has been made, critical barriers persist, including data scarcity, the black-box nature of deep learning (DL), and the complexities of scaling lab-scale insights to industrial systems. Addressing these gaps through standardized data protocols and explainable AI (XAI) is essential for the deployment of dependable, economically viable hydrogen technologies.