Because of the rapid increase in both photovoltaic and concentrating solar power plant, the request of very accurate direct and global radiation forecast is fast growing too. A good forecast accuracy, for example, allows the Transmission System Operator (TSO) to enhance the planning and management of the energy reserve. On the other hand, for the producer, a good performance in the forecast may increase the economic value of solar power in the day ahead energy market. A big issue, in the solar power forecast, is the cloud variability and the ability of the meteorological models in reproducing it. In order to validate solar radiation forecast, accurate measurements of the different components of solar radiation are needed, but an important issue is also to know the confidence interval of the forecast for different sky conditions: sunny, partly cloudy or cover. In particular, it would be of great interest to evaluate the model performance in presence of specific types of clouds. It is known, in fact, that each type of cloud presents different optical properties, not always well described by the models, influencing the incoming radiation. Ground based instruments, such as ceilometers and total sky imager, that could be used for this purpose, are not so widespread and therefore it’s not possible to verify the forecast over wide areas. This is one of the issues which drove our research on the use of satellite data to detect the cloud type. The SAFNWC software has been implemented and adapted to the particular input format of HRIT Meteosat Second Generation images. The software allows to characterise twenty different types of clouds on the basis of multispectral analysis from the SEVIRI radiometer data.In order to evaluate the influence of the cloud type on the measured ground irradiation, a preliminary statistical analysis of these two variables has been performed on three Italian sites.Then the error on irradiation forecast, obtained from a radiative transfer model, nested into a Numerical Weather Prediction (NWP) model, has been analysed for different illumination and cloud type conditions. The used forecast system has been figured out implementing different cloud schemes, using either NWP cloudiness or a diagnostic radiative cloudiness, obtained from vertical humidity and some tunable thresholds. The results have been analysed on three sites on the North, Centre and South of Italy, differentiating results on the basis of the different types of clouds according to SAFNWC.