Application to photovoltaic forecasting using ground images
Physics-inspired machine learning - Principles and application to solar energy forecasting

The share of renewable energies in the energy mix has risen sharply in recent years. However, the intermittency of their production remains a real challenge for their large-scale integration into existing power grids. The grid operator must ensure that electricity production and consumption are balanced at all times. The challenge also lies in the independent control of photovoltaic or wind farms, which can be coupled with additional storage or production resources, particularly in isolated island systems.

In this context, EDF R&D has been working for several years on photovoltaic production forecasts, for different time horizons and using different input data (meteorological models, satellite images, ground images, real-time...