Théo DELFOLIE

__________________________________________________________________________________

Personal information

E-mail:

theo.delfolie@univ-smb.fr (LISTIC)

theo.delfolie@cea.fr (CEA)

Office:

A224

Address:

LISTIC, 5 chemin de bellevue, CS80439, 74 944 Annecy Cedex

Reaserch team: 

AFuTé

Links:

Linkedin – GitHub

__________________________________________________________________________________

Thesis Information

Topic:

Solar Panels Soiling Prediction via real-world environment modeling and data fusion

Keywords:

Artificial Intelligence – PV Soiling Losses – Data Fusion – Solar Energy – Sustainability

Abstract:

Photovoltaic (PV) power plants, especially those located in areas prone to soiling – such as arid regions, coastal environments, and agricultural sites – can experience energy losses of up to 20-30% per year. In 2023, this translated into financial losses exceeding 10 billion euros. This PhD project aims to develop a robust and comprehensive method to detect and predict soiling losses on PV systems. The approach combines real-world environmental modeling, operational data (electrical, thermal, optical), and the use of data fusion techniques, artificial intelligence (AI), and multispectral imaging. The research will follow a bottom-up approach structured in three main stages: 1. Component/module level: Laboratory-based reproduction and modeling of soiling accumulation under controlled conditions, followed by experimental validation. This stage will leverage CEA’s expertise in degradation mechanisms and accelerated testing protocols. 2. Module/system level: Launch of measurement campaigns to collect diverse datasets (meteorological, operational, imaging) and conduct field tests on a pilot site. These data will be used to validate and enhance CEA’s diagnostic tools by incorporating innovative features such as AI-driven soiling losses prediction and forecasting. 3. System/operational level: Validation of the method on commercial PV modules deployed in actual PV power plants, with the goal of demonstrating its scalability and real-world applicability. The outcomes of this PhD will contribute to the development of an innovative tool/method for comprehensive soiling diagnosis and prediction in PV installations. This will help minimize energy losses and support the anticipation and optimization of cleaning strategies at scale for PV power plants.

Start of the Thesis:

December 2025

Doctoral School:

Sciences, Ingénierie, Environnement (SIE)