Maintaining a constant voltage in polymer electrolyte membrane fuel cells (PEMFCs) has always attracted the attention of many researchers, and many articles have been published on this issue. Furthermore, water management in PEMFC has become an important challenge because it can improve cell efficiency and lifetime. This paper will develop a one-dimensional dynamic model for a single PEMFC, which correlates changes in the cell voltage to changes in the cell current density and humidification rate. Subsequently, a recurrent neural network controller based on the approximation of nonlinear autoregressive moving average model is proposed. The controller manipulates the anode and the cathode water mole fractions in order to fix cell voltage and preserve cell water content within a satisfactory interval regardless of the varying cell current. The model and the controller are simulated in matlab/Simulink (Mathworks Inc., Natick, MA) software, and the results are compared with a PID controller from different viewpoints such as current disturbance and plant parameter variation. Copyright (c) 2011 John Wiley & Sons, Ltd.