Wind energy predictions have been widely developed with a variety of methods, this is due to the stochastic character and uncertainty in the wind. The need for wind energy generation is so great that it must be prepared for operational prediction on its network. This study is very important that aims to design an algorithm to predict wind power for grid operators that are useful to accelerate the management planning of the generation so that the resulting wind power is more optimal. In this paper, we propose a model of wind power prediction by attaching highly intermittent wind speed behavior that makes wind power change rapidly. To overcome this, Wavelet Decomposition method is proposed, then this model is hybridized using Nonlinear autoregressive using Nonlinear autoregressive modeling machine with exogenous input model Nonlinear Autoregressive with External-Neural Network (NARX-NN). The simulation results show that this model can improve the accuracy performance of previous models using BP-Neural Network.