In this research article, it has been implemented an spatio-temporal active power (AP) forecast based on the Kriging theory and Adaptive wavelet neural network (AWNN) by using Julia Programming; it considers the wind speed (WS) characteristics of highly stochastic and random features with non-stationary data, with data calibrated with 21 years of data (2000 to 2021); it is considered with the influence; the physical model is structured by Kriging theory for the wind speed at hub height, according the manufacturer curve in the wind farm, the model is a input in the statistical model for the active power forecast. Our findings are the improved accuracy compared with the ARX 72.4%, ARMAX 75.5% and fuzzy 81.1% approaches, by using spatio-temporal wind forecasts, the accuracy is increased as 89.2%.