In this work, we aim at perform short-term prediction on the solar irradiance in order to allow the Photovoltaics (PV) operators to manage and allocate energy supplies. Making accurate short-term prediction helps ensure the stability of power supply without the need to reserve too much energy. All-sky cameras are devices that can be used to monitor the behaviors of the sun and the clouds. By analyzing all-sky images, we extract features that can be used to predict solar irradiance via a trained regression model. The results of this work could provide very useful information for PV operators to ensure greater efficiency of the solar energy supply. Categories and Subject Descriptors I.5.m [Pattern Recognition]: Miscellaneous General Terms Algorithms.