Using Sentinel data in CNN to automatically identify solar power plants in Italy: a comparison of different spectral band combinations

Interest in space data applications has been increasing in recent years due to radical digitalization and technological transitions. Satellite data is considered an innovative data source that contributes to the development of models and decision-making policies on relevant topics, including sustainability and energy-related concerns. In our study, we present an implementation of Sentinel-2 data, which is freely available from the European Space Agency’s Copernicus project. We integrate this data into a Faster R-CNN algorithm with the goal of automatically detecting solar power plants. We also compare different spectral band combinations to enhance the algorithm’s classification ability in terms of average precision. This paper presents an automated approach to detecting solar power plants within Italian territory. It contributes to providing estimates of the growth of renewable and clean energy in the country, while also generating insights to facilitate informed and optimal decision-making.