The prediction of the electric energy production by solar energy systems can be used as a real time monitoring of its performance, allowing the identification of operational problems or its performance degradation, as well as the short and medium term production of electric energy, thus optimizing the use of the electric network. Normally, the daylong climactic data prediction is useful for helping stabilizing the electric network. The one or more day ahead prediction of electric energy is a useful tool for planning the energy supply to the market. The main goal of this research is to propose a short-term prediction model for the solar irradiation by means of weather condition correlations, in doing so, we expect to lower the uncertainties of the solar irradiation forecast, for which can be used to forecast photovoltaic energy. The proposed prediction model uses literature available correlations for clear sky (no clouds) and overcast days, to predict the solar irradiation in other days, through the use the ambient temperature and atmospheric transmission literature available correlations. Several prediction models were used and combined for 1, 3, 5 and 7 days ahead prediction of the available average solar energy in that time interval . It was concluded that the clear sky model plays an important role in the prediction reliability. Experimental data taken every minute for almost four years in the city of Miranda do Norte in the state of Maranhão, Brazil, were used to verify and propose a model to better predict the available solar energy irradiation, to within 22-24 percent, 14-16 percent, 13-15 percent and 12-13 percent uncertainty (95,45 percent confidence level), respectively for 1, 3, 5 and 7 days ahead of the available solar irradiation in the time interval. The least square model was also analyzed and it was shown to predict the solar irradiation with much higher uncertainties. As a result, the proposed model can be easily implemented for predictions in other places and can be very useful for predicting and planning the daily and weekly output of photovoltaic plants. In this dissertation the prediction model suggested by Bindi was utilized, together with three clear sky models to reduce the prediction uncertainty.