This thesis is to develop an artificial neural network (ANN) model of a single-cell PEM fuel cell test station for predicting the relationship of input variables and output variables. The architecture of neural network is consisted of 4 nodes of input layer, 4 nodes of a hidden layer with tanh-sigmoid activation functions and three nodes of output layer with linear activation function. Specify the cell temperature, flow of hydrogen, flow of oxygen and current density as inputs, with water production, cell voltage and power as outputs. To train the network, Levenberg-Marquardt back propagation (LMBP) was used since it is an efficient method. The solution was quickly converge and better target was achieved. The obtained ANN was used as a process model for designing a control system. The system was tested and its performance was satisfactory.