PEM fuel cell system modeling is a potential tool to design, analyze and predict its performances in different operating conditions. Many modeling approaches such as CFD, electrical equivalent circuit, empirical, semi-empirical and lumped parametric models have been applied to simulate PEMFC systems from the level of the MEA (membrane-electrodeassembly) unit cells up to full stack applications. In this paper, a new approach for a 5kW PEMFC system modeling is proposed. The FC system exhibits a nonlinear behavior captured from its experimental polarization curve. The latter is linearized around specific operation points. At each operating point, a local linear identified model that describes the behavior of the FC system at specific working conditions is developed. A data driven model based on Takagi Sugeno Fuzzy (TSF) approach is used to predict the global nonlinear behavior of the FC system for given operating conditions and driving inputs. In a controlled environment (stack temperature, humidity, and reactants pressure) the considered uncontrolled input for the system is the current load and the output is the voltage and electric power. The response of the proposed model is evaluated for different profiles of the load current. The developed model predicts the transitory behavior of the FC system accurately with a relative error of 0.4% and with a significant reduction in computational time making it suitable for integration and real time simulation of FCEV powertrains.