Description du sujet de thèse

-Context: The energy transition requires a massive and efficient deployment of renewable energies, among which rooftop photovoltaic (RPV) systems play a central role. However, the lack of precise information on the location, size, and technical characteristics of existing RPV systems severely limits the ability to plan their development and optimize their integration into the power grid.
This PhD focuses on developing an innovative methodology to accurately locate and characterize rooftop photovoltaic installations using advanced artificial intelligence (AI) methods, combined with diverse geospatial data (aerial imagery, LiDAR, cadastral data, socio-economic data).

PhD Objectives:

Build a unified database integrating detailed spatial information (high-resolution aerial images, LiDAR, cadastre) together with socio-economic and technical data on buildings.

Develop and validate deep learning classification and segmentation models at the building scale to precisely identify installed photovoltaic systems.

Explore the use of federated artificial intelligence to effectively manage the spatial and temporal heterogeneity of data, thereby ensuring model robustness and adaptability.

Provide accurate and scalable interactive maps to analyze the spatial distribution of rooftop photovoltaic systems and study the dynamics and factors influencing their adoption.

Methodology: The candidate will combine deep learning approaches such as convolutional neural networks (CNNs) and semantic segmentation architectures (U-Net). Particular attention will be given to federated AI methods to enhance model performance on geographically distributed and heterogeneous datasets. The work will rely on existing and open-access databases, as well as advanced geospatial tools to ensure accuracy and transferability of the results.

Expected Results:

Development of a reproducible and robust methodology for identifying RPV systems.

Production of models with large-scale generalization capabilities.

Open-access dissemination of datasets and developed tools.

Scientific dissemination through publications in international journals and presentations at major conferences in the field.

Contexte de travail

Working Environment:
This PhD will be carried out at the LOCIE laboratory (UMR 5271 CNRS – Université Savoie Mont Blanc, Institut National de l'Énergie Solaire). The research will take place at LOCIE, 60 avenue du Lac Léman, Savoie Technolac, 73376 Le Bourget-du-Lac.

USMB (Université Savoie Mont Blanc) – With 15,000 students, a rich and multidisciplinary range of academic programs, and 18 internationally recognized research laboratories, Université Savoie Mont Blanc (Chambéry) is a high-level research and multidisciplinary university that has developed significant expertise in solar energy, building energy efficiency, as well as modeling and information sciences.

CNRS (French National Centre for Scientific Research) is one of the world's leading research organizations. Its researchers explore the living world, matter, the Universe, and the functioning of human societies in order to address today's and tomorrow's major challenges. Internationally recognized for the excellence of its scientific research, the CNRS is a global reference in research and development as well as for the general public.

INES (National Institute of Solar Energy) is a global leader in R&D, expertise, and training in advanced solar technologies, their integration into systems, and smart energy management.