The need to have a tool capable of predicting with good accuracy the level of solar radiation on the ground comes from the activity of balancing the electricity grids by the operators. In particular, the energy authorities require each energy producer to provide a daily estimate of the amount of energy that will be produced. In a free energy market context, any imbalance between declared energy and energy actually produced can create serious problems for producers.
In fact, the grid operator will force the producer to pay for the purchase of the energy that has not been produced, but at the same time it may not pay for the surplus energy if it is not necessary to balance the grid. Hence the importance of a short-term forecasting system for the energy produced. While this estimate can be easily obtained for non-renewable sources, it becomes extremely difficult for unstable sources such as wind and photovoltaic. The startup Sol-Invictus proposes an innovative forecasting system based on the synergistic use of meteorological satellite images and deep learning algorithms.