This paper presents a new statistical short-term forecasting system for a grid-connected photovoltaic (PV) plant. The proposed system comprises three modules composed of two numerical weather prediction models and an artificial neural network based model. The first two modules are used to forecast weather variables used by the third module, which has been selected from a set of different models. The final forecast value is the hourly energy production in the PV plant. The forecasting horizon ranges from 1 to 39 h, covering all of the following day. The forecast values can be used for determining the most favourable hours to carry out maintenance tasks in the plant, and for preparing bid offers to the electricity market. (C) 2012 Elsevier Ltd. All rights reserved.