By combining detailed knowledge acquired from the established off-line characterisation of fuel cells with a careful categorisation of signatures of faults from on-line measurements, it may be possible to develop a complex and accurate understanding of the health state of electrochemical systems from more simple sensing when advanced machine learning and data-driven approaches are taken.

Abstract

Polymer electrolyte fuel cells (PEFCs) are regarded as a substitution for the combustion engine with high energy conversion efficiency and zero CO₂ emissions. Stable system operation requires control within a relatively narrow range of operating conditions to achieve the optimal output, leading to faults that can easily cause accelerated degradation when operating conditions deviate from the control targets. Performance recovery of the system can be realized through early fault diagnosis; therefore, accurate and effective diagnostic characterisation is vital for long-term serving. A review of off-line and on-line techniques applied to the fault diagnosis of fuel cells is presented in this work. Off-line approaches include electrochemical impedance spectroscopy (EIS), cyclic voltammetry (CV), galvanostatic charge (GSC), visualisation-based and image-based techniques; the on-line methods can be divided into model-based, data-driven, signal-based and hybrid methods. Since each methodology has advantages and drawbacks, its effectiveness is analysed, and limitations are highlighted.