The aim of this work is to develop and deploy an advanced model-based control framework for a polymer electrolyte membrane (PEM) fuel cell system. The framework relies on nonlinear model predictive control (NMPC) using a reliable and efficient dynamic optimization approach which discretizes both manipulated and state variables. The optimization is performed using a direct transcription method that handles the optimal control problem as a nonlinear programming (NLP) problem. The motivation for the control is to ensure optimum power generation following a variable load demand with acceptable response time while avoiding oxygen starvation and minimizing hydrogen consumption. To validate the applicability, efficiency and robustness of the NMPC scheme a small-scale fully automated unit was used and an experimentally validated semi-empirical dynamic model was utilized at the core of the optimization scheme. The on-line application of the multivariable controller shows that the proposed framework can accomplish the desired objectives for power demand in the context of a safe operating region. Furthermore the controller exhibits excellent performance in terms of computational requirements and can follow load changes with a negligible error in its response, even at varying operating conditions. (c) 2013 Elsevier Ltd. All rights reserved.