Reduced-dimensionality nonlinear distributed-parameter observer for fuel cell systems

Abstract

To ensure reliable and efficient operation of fuel cell systems, it is important to monitor them online. However, placing sensors inside the fuel cell is often challenging, so virtual sensing using an efficient state observer is used in this study. Detecting local internal phenomena, such as reactants’ starvation, membrane dryout/flooding, and nitrogen accumulation, requires knowledge of the spatial distribution of internal states. Lumped-parameter models are not suitable for this, as they use a single variable to describe parameters such as hydrogen concentration. Instead, a high-order distributed-parameter fuel cell model is used to predict the spatial profiles of various internal states. An observer algorithm is employed to correct the predicted quantities using a few measurements taken at the system boundary. This update step only considers dominant dynamics from a reduced model to adjust all system states accordingly, making it computationally efficient and robust. The observer algorithm’s performance was verified against a high-fidelity model through detailed simulations.

Keywords

• Fuel cell
• Balanced truncation
• Distributed state estimation
• Extended kalman filter