Joint project BZB

 

Real-Time Robust Estimation of Stack Internal States for PEM Fuel Cell Systems

 

Motivation

Fuel cell systems can contribute to a decarbonization of future mobility solutions. While steady-state operation of fuel cell systems is already implemented in various systems, the main challenges for extensive use in e.g. fuel cell vehicles besides system costs are control algorithms for transient yet safe operation. Violating operational constraints can degrade the performance or even permanently damage the fuel cell and must be avoided. Certain internal states, such as water and temperature distribution within the fuel cell, are found to be critical for maintaining safe operation. Yet those internal states can rarely be measured directly, especially not in a series production setup. Developing advanced estimation algorithms for those internal states as software sensors and utilizing the knowledge in the control algorithms overcomes this dilemma and can accelerate the future use of fuel cell systems.

 

Goals and Methodology

For the development of the aforementioned estimation framework, we use a model-based approach. For that purpose, high-fidelity models and measurement data are utilized to identify and analyze the behavior of the critical states to be estimated under various relevant operating conditions. This information is then used to derive a model which offers a good trade-off between model accuracy and computational complexity and is thus suitable for real-time execution and implementation of the observer algorithms. The critical internal states should be observable through this model and, ideally, easy-to-access measurements, i.e., available sensors. Different model-based observer algorithms will be implemented, and the overall framework evaluated concerning required sensors, estimation accuracy and robustness as well as computational efficiency. For the latter, the algorithms will be tested on RCP hardware.

 

Innovations and Future Use

The implementation of the project will provide a real-time capable virtual sensor whose insights into critical non-measurable states of the fuel cell will enable significant opportunities for new real-time capable control strategies of fuel cell systems. These are mainly aimed at improved performance with high dynamic response while maintaining high durability of the fuel cell. This will be a further step towards the wide use of fuel cell systems in mobile applications.

 

Project partner
Ford Research & Advanced Engineering