In this paper, we propose an innovative control strategy combines direct power control (DPC) with artificial neural network (ANN) applied to Doubly-Fed Induction Generators (DFIG) used in variable speed wind energy conversion systems. The conventional DPC, based on a fixed switching table and hysteresis controllers, remains sensitive to rotor speed variations and its performance degrades under non-ideal conditions, such as wind speed variations and grid disturbances. To overcome these limitations, we have combined DPC with neural networks (DPC_ANN) to replace the traditional switching table and to predict the optimal switching states based on dynamic speed variations. A model of a 1.5 KW DFIG is developed using Matlab/Simulink to test the proposal approach. Several test scenarios are investigated, including variable wind profiles and parametric uncertainties. The results show that the DPC_ANN reduces active power and reactive power oscillations and improves the transient performance compared to the classical DPC. Moreover, the total harmonic distortion (THD) rate was reduced. These studies confirm that DPC_ANN strategy offers enhanced robustness under varying conditions, paving the way for increased reliability and efficiency in Wind Energy Conversion Systems (WECS).