The availability of reliable data related to the behavior of the wind in a wind farm is very useful to determinate with accuracy some aspects as power curve of a wind energy turbine and the wind's speed in an interval of times. The precision in the prediction process of wind behavior is useful for reducing structural efforts in the wind-turbine rotor system, and even in the tower section. In practice, sensors are used to acquire data for monitoring wind farms which occasionally may tend to fail causing an incomplete information from the sensor. This work is focused on the imputation of missing data based on a combination of interpolation and regression models. Our experiments show that this approach is useful for correlated time series, considering the direction and speed wind for 20 and 40 meters of height in wind farms located in the Isthmus of Tehuantepec in Oaxaca state in Mexico. Finally, the approach of combination methods is effective to solve the problem of missing data in the database of wind in wind farms.