The paper presents the control of a doubly fed induction generator (DFIG) connected with a three-level neutral point clamped (3L-NPC) inverter with. Model Predictive Direct Power Control (MPDPC) is synthesized using a dynamical model of the DFIG and 3L-NPC inverter. The principle of the proposed control scheme is to use the dynamical model to calculate predictions of the future values of the stator flux, rotor current and DC-link capacitor voltages for all possible configurations of voltage vectors. The active and reactive power can be estimated based on the stator flux and the rotor current. A cost function will be used to obtain the predicted profile which minimizes the error between the active, reactive powers and their references, balance the DC-link capacitor voltage and reduce the switching frequency and common-mode voltage. The optimal switching state that minimizes the cost function is selected and applied to the inverter. Simulation results under different conditions of wind speed are presented and compared with Deadbeat power control and Space Vector Modulation (DPC-SVM). The obtained results show the improved performances of the proposed control method.