Successful internships

Have you ever wondered how weather forecasts are made? Sam Ghyselen, student at IMT Atlantique, has worked on that subject during his internship at HYGEOS. He has developed a model to predict cloud movement and properties from Meteosat images. He conducted researches on methods used for images time series analysis and prediction. More and more papers are published to address this subject in the meteorology field, providing for this mission a solid scientific background.  So, he managed to implement a specific architecture of an advanced artificial intelligence model. He focused on the cloud optical thickness prediction, as this parameter is the main cloud property used by SolaRes, a state-of-the-art solar resource estimator developed by HYGEOS. The model is shown to be accurate for predictions leading up to 3 hours, paving the way for potential features of SolaRes like solar resource prediction.

Cloud Optical Thickness prediction for 13^th July 2024.

In parallel, Pierre Posson, student at ESTACA, designed an electrical conversion module enabling the transformation of solar irradiance into photovoltaic power output. This module was tested on the L2EP ENSAM Lille PV installations, with results compared against both in-situ measurements and the PVGIS online tool. He also developed an automated tool for the real-time or historic comparison of SolaRes outputs with ground-based observations, but also with other reference products. This approach strengthens the validation of SolaRes performance while highlighting its added value compared to existing methods, and establishes a framework for its continuous improvement by integrating feedback from both observations and models. These developments make it easier to communicate with end-users, by translating solar resource estimation directly into electrical production.

Validation of the full conversion chain up to alternating current (AC) power production.