The importance of methods using renewable resources for the energy production has increased drastically in the recent years. In the electricity market the method that has grown most rapidly is the generation of power through wind. Forecasts of this power production are crucial. However, difficulties arise in these forecasts due to the high volatility of the wind and the complexity of the terrain where the wind facilities are installed. This thesis examines five different models for the forecast of wind energy and analyses the behaviour of its errors. The scope is to build a confidence interval for the errors of these forecasts. In this framework two major setups are considered for constructing a confidence interval. The first one is based on applying a GARCH model in order to forecast the conditional variance of the errors. The second uses a regression model to describe the absolute or square errors as a function of explanatory variables like the predicted wind speed and predicted wind power. These models were then used to build the confidence intervals. Both approaches were implemented sucessfully.