Polybenzimidazole fuel cells are a relatively recent type of fuel cells that operates above 100 degrees Celsius. Because of their high temperature, catalytic activity is increased, along with tolerance to CO in the feed. Furthermore, water management is not an issue any more, as water is produced completely in gaseous form.

To provide a framework for development of a control system for PBI fuel cells, a dynamic model of such a fuel cell has been developed. Many models in the literature, both steady-state and dynamic ones, consider the current density as input. However, this is not realistic from the point of view of control engineering, because the current density is hardly a manipulated variable, since it is determined by the external circuit's characteristic, and not directly by the operator. Instead, this model uses a generic, time-dependent external characteristic as input; the characteristic can in turn be modified in real time by e.g. manipulating the gate voltage on a MOSFET transistor, which is a common way of controlling fuel cells. Control of such a system presents difficulties in that it contains strong nonlinearities, and dramatic changes in system properties with some variables as temperature and CO content of the feed. Once implemented, the model has been able to simulate the transient patterns observed on voltage vs. current density diagrams in laboratory experiments, which follow (for the case of a step change in a purely resistive load) two straight lines in the voltage vs. current density diagram; the model has also reproduced another typical feature observed in experiments: the relaxation times in transients are often different in apparently symmetric situations, as for example step-changing a resistive load to a differently sized resistance, and then stepping back to the original one; more generally, it explains why it could be possible to observe relaxation patterns made of a sequence of a step, a(n almost) linear part, and a final exponential part.

The theoretical analysis has, among others, the interesting result that, in theory and under sufficiently general assumptions on operational conditions, perfect control of the fuel cell's power output is possible, since there is always an overshoot in the power output in connection with a step change in the input variable.

The model has been originally been developed in Scilab, but is now in the process of being ported to SimuLink for controller design.