The unstoppable spread of photovoltaic production drives the search for intelligent solutions to mitigate imbalances between energy supply and demand, manage grid instability and ensure stable profits for the producer. Due to the development of energy markets with multiple time periods, there is an increasing need for production forecasts on multiple time scales, from fifteen minutes up to days ahead. To address this problem, this study presents both a short-term (three days ahead) and a very short-term (three hours ahead) PV production forecasting method. The short term is based on a multi-model approach and refers to different configurations of the Analog Ensemble method, using the forecasts of four numerical weather forecasting models. The very short-term one consists of an Auto-Regressive Integrated Moving Average Model with eXogenous input (ARIMAX) which uses the short-term power forecast output and irradiance from satellite processing as exogenous variables. The methods—applied for one year to four small grid-connected plants in Italy—achieved promising improvements compared to reference methods. The study also revealed the usefulness of satellite cloud cover data to interpret correctly the results of performance analysis.