Machine learning utilized for the development of proton exchange membrane electrolyzers

Proton exchange membrane water electrolyzers (PEMWEs) have great potential as energy conversion devices for storing renewable electricity into hydrogen energy. However, their cost and efficiency are still unable to support large-scale commercialization. In PEMWEs, multiple processes at different scales are simultaneously involved: electrochemical reactions at the microscale, reactant transportation at the mesoscale, and thermoelectric-electrical coupling fields at the macroscale. Therefore, the system is so complex that previous studies could only focus on the optimization of different subsystems of PEMWEs and merely investigate limited variables separately: electrocatalysts, membrane electrode assemblies (MEAs), single cells, stacks, etc. Therefore, the contributions of traditional research have been limited, and the trial-and-error method adopted to drive the experimental or theoretical studies is inefficient. Reports of applying machine learning (ML) to accelerate the research and development of key materials in PEMWEs have gradually emerged in recent years. The application of ML can greatly accelerate the optimization of different key materials and components in PEMWEs, significantly reducing unacceptably high costs. Therefore, this paper reviews recent applications of ML for material and component optimization in PEMWEs. Moreover, we provide a phased summary and prospects for the future of this emerging field.