This brief presents an unknown input observer design method for a class of nonlinear systems in the presence of disturbances in both the system dynamics and the output. The theoretical result proposed can be considered as an extension of those established in [1] and [2], which are dedicated to observer design for estimating both the state of the system and an unknown input. By means of regular transformations, the robust observer design problem is turned into a linear matrix inequality feasibility problem. One of the main motivations of this brief is to develop a state estimator for proton exchange membrane fuel cell stack systems. More precisely, this brief addresses the problem of observer design for estimating the main variables of the fuel cell systems. Simulation results are presented to highlight the good convergence properties of the observers proposed.