A comparison of three quasi-LPV models of a PEM fuel cell used for Nonlinear Model Predictive Control

This paper is concerned with the power control of a PEM fuel cell system with integrated DC/DC converter using Nonlinear Model Predictive Control. For the controller design three distinct approaches to model the system in quasi-Linear Parameter Varying representation are examined. Thereby, all nonlinearities of the system are expressed by scheduled parameters by which the model of the plant becomes linear in its states and inputs. Two of the approaches utilize highly nonlinear terms, whereas the third approach is based on scheduled parameters set up in a 1-dimensional lookup table. Finally, the performance of the evolved predictive controllers using distinct models of the plant are compared.