Enhancing Solar Radiation Conversion in Floating Photovoltaic Systems Through Machine Learning

Overreliance on traditional energy sources leads to resource depletion and environmental pollution, posing serious challenges to global energy security. In response, Floating Photovoltaic (FPV) technology has emerged as a promising solution, offering an alternative to traditional ground-mounted Photovoltaic (PV) systems. This study utilizes data collected from Brookhall Estate between September 2022 and January 2023, employing six machine learning algorithms and the interpretable SHapley Additive exPlanations (SHAP) technique to analyze solar radiation convertibility. Random Forest is identified as the most reliable model for FPV prediction. The results highlight the nonlinear impact of factors such as temperature, conductivity, seasonal variations, water depth, and wind direction on solar radiation collection efficiency. Targeted strategies for FPV design and placement are proposed, contributing to overall FPV optimization in alignment with sustainable energy development goals and the pursuit of net-zero emissions.