Prescriptive Lifetime Management for PEM fuel cell systems in transportation applications, Part II: On-board operando feature extraction, condition assessment and lifetime prediction

Abstract

The 2030 and 2050 targets for lifetime requirements of proton exchange membrane (PEM) fuel cell (FC) systems in heavy-duty vehicle (HDV) applications are set to 25,000h and 30,000h, respectively. The need to reduce PEM fuel cell system (FCS) costs is expected to be accompanied by greater uncertainty in lifetime. This will require robust and reliable approaches and algorithms for feature extraction, condition assessment and lifetime prediction to obtain the Remaining Useful Life (RUL) and to prevent premature failure, downtime and lower system availability leading to higher total cost of ownership (TCO).Current approaches to feature extraction and component-based condition assessment typically focus on in/ex situ measurement methods performed on test benches, rather than operando measurement methods that can be performed on-board commercial vehicles. Methods developed on test benches are often difficult or even impossible to transfer to commercial vehicle applications. As a result, the lifetime prediction in commercial vehicle applications often focuses on data-driven approaches to predict voltage or power decay. Since these data-driven approaches do not contain sufficient information about the component-based State of Health (SoH), a component-based lifetime prediction and subsequent optimization of operating strategy to control RUL and meet the lifetime requirements is significantly limited.Prescriptive Lifetime Management is a technique that involves developing and implementing on-board strategies and algorithms for feature extraction, condition assessment and lifetime prediction to mitigate degradation, extend the RUL and reduce the overall TCO, e.g., of the PEM fuel cell system through global optimization-based post-prognostic decision-making. With this in mind, this paper presents approaches for on-board operando feature extraction and condition assessment of the PEM integrity and the electrochemical surface area (ECSA). The calibration of macroscopic degradation models allows for lifetime prediction and RUL control to meet the lifetime requirements of PEM fuel cells. The results show that the operating parameters have a high effect on PEM fuel cell degradation. The complexity of degradation prevents the generation of reliable degradation lookup tables. Therefore, the direct implementation of calibratable macroscopic degradation models in the Prescriptive Lifetime Management framework is recommended.