Model-Based Algorithm for Water Management Diagnosis and Control of PEMFC Systems for Motive Applications

Abstract

Water management in PEMFC power generation systems is a key point to guarantee optimal performances and durability. It is known that a poor water management has a direct impact on PEMFC voltage, both in drying and flooding conditions: furthermore, water management entails phenomena from micro-scale, i.e., formation and water transport within membrane, to meso-scale, i.e., water capillary transport inside the GDL, up to the macro-scale, i.e., water droplet formation and removal from the GFC. Water transport mechanisms through the membrane are well known in literature, but typically a high computational burden is requested for their proper simulation. To deal with this issue, the authors have developed an analytical model for the water membrane content simulation as function of stack temperature and current density, for fast on-board monitoring and control purposes, with good fit with literature data. The water flow from the catalyst layer to the GFC through the GDL is modelled considering as main transport mechanism the capillary transport. The water coming from the GDL then emerges through the pores inside the channel forming water droplets that interact with the air flow. The authors have developed several papers on this topic: mathematical models have been developed for droplet’s emersion, oscillation, and detachment phases; furthermore, the coalescence between near droplets has been included into the modelling. The authors have also validated with experimental results the proposed models. The objective of this paper is to develop a mathematical model able to represent a typical fuel cell stack in order to predict the water membrane content and the water removal rate, that are fundamental to correctly control the PEMFC system in order to avoid the critical conditions mentioned before, ensuring the best performances of the stack reducing the hydrogen consumption. The model is validated with literature data, showing optimal fit and high correlation, making it suitable for further analyses.