Real-time machine learning-driven digital twin framework of a floating solar system in waves

Abstract

Floating photovoltaic (FPV) systems offer a promising pathway for sustainable energy generation by avoiding land occupation, and utilising water cooling to improve energy efficiency. However, their deployment in dynamic aquatic environments introduces significant challenges, including irradiance fluctuations, hydrodynamic loads, and structural fatigue, which complicate reliable performance prediction. These highlight the need for advanced tools that can enable offshore structural monitoring, maintenance decision-making, and power management in a dynamic environment. To address these challenges, this study introduces a real-time digital twin framework for FPV systems, integrating laboratory experimentation, machine learning, and real-time visualisation. A bespoke facility combining a solar simulator, FPV prototype, and wave tank enabled 155 controlled tests under diverse irradiance and wave conditions. The resulting dataset was used to train a Random Forest model, which achieved an overall coefficient of determination of 0.990 and accurately predicted the dynamic responses of heave, surge, and pitch, with minimal discrepancies. The results were visualised through a real-time Unity-based user interface, enabling intuitive interaction and monitoring of the FPV system’s behaviour. These findings demonstrate the potential of AI-enabled digital twins to enhance operational resilience, reduce maintenance costs, and pave the way for intelligent, adaptive control of large-scale offshore FPV deployments.