Reuniwatt is excited to announce its participation in the European Geosciences Union (EGU) General Assembly 2026, where we will present our latest work on photovoltaic soiling losses. Join us for our poster session on 📍<\/span>Monday, 04 May 2026, from 16:15\u201318:00 (CEST) at Hall X4, Display time from 14:00\u201318:00. Soiling losses are a significant source of uncertainty in photovoltaic energy yield calculations. This issue is especially prevalent in regions that experience high aerosol concentrations and intermittent precipitation, such as West Africa and the Pacific Islands. Soiling is influenced by complex meteorological factors, including deposition processes, rain events, wind-driven dust transport, and proximity to local aerosol sources, making it challenging to estimate accurately.<\/p>\n For our study, we analyzed one year of field measurements from monitoring networks in West Africa and the Pacific Islands to assess soiling losses. By integrating meteorological drivers from ECMWF reanalysis, including precipitation and particulate matter, our team examined the performance of two well-established semi-physical soiling models, HSU and Kimber. We eagerly anticipate engaging with fellow researchers and industry professionals in Vienna and invite all attendees to visit our poster session to explore the details of our hybrid model for soiling loss estimation!<\/p>\n<\/div><\/div><\/div><\/div><\/div>
The session will feature our presentation titled Hybrid physical-machine learning estimation of photovoltaic soiling losses from meteorological reanalysis data in Africa and Pacific islands<\/em>. Join us to learn how machine learning and physical models can improve the estimation of energy yield losses caused by soiling.<\/p>\n![\"Soiling](\"https:\/\/reuniwatt.com\/en\/wp-content\/uploads\/sites\/3\/2026\/04\/PV_Soiling_Losses_Hybrid_Framework-600x338.jpg\" "\\"\\"")
<\/p>\nSoiling Losses in Photovoltaic Systems<\/h2>\n
Hybrid framework for improved estimation of soiling losses<\/h2>\n
We then introduced a hybrid framework, which demonstrates improved soiling loss estimation across most regions, surpassing the performance of the traditional models. However, it is less effective in areas with low soiling levels. Our poster highlights the potential and limitations of hybrid physical-machine learning approaches for meteorology-driven soiling assessment. The hybrid model developed in this study not only offers more accurate soiling loss estimations but also supports practical applications, such as optimized maintenance decisions and improved photovoltaic energy yield forecasting. By integrating both physical models and machine learning, we can better understand the impact of meteorological factors on solar panel performance and contribute to more efficient energy production in areas affected by soiling.<\/p>\n