Model based predictive control originated in the late seventies and has been developed considerably since then. Predictive control is now widely regarded as one of the standard control algorithms for industrial processes. In this thesis a new approach to design generalized predictive controller for nonlinear systems with model uncertainty is proposed. The controller is based on a state-space model composed of generalized predictive control, uncertainty compensator and feedback linearization. The uncertainty compensator is proposed in this method to remedy the drawbacks of the model based algorithm used in both GPC and Feedback Linearization structure. Stability analysis of the method is elaborated and a comparison study with of the nonlinear predictive methods is performed using computer simulations. The efficiency of this method is illustrated through simulations.