Prediction of floating solar power plant mooring dynamics via machine learning techniques

• Machine learning predicts mooring loads and motions in floating solar plants. • Gradient boosting outperforms other methods for tension and displacement prediction. • Random forest gives better estimates of platform yaw rotation behavior. • Data-driven models reduce simulation time from hours to seconds. • Real case study shows benefits for floating solar plant maintenance and design. The rapid expansion of solar photovoltaic technology has highlighted the potential of floating photovoltaics (FPV) to address land use limitations while providing additional environmental benefits. However, maintaining critical mooring systems that ensure FPV stability remains challenging, especially under environmental loads such as wind and waves. This study explores the application of machine learning (ML) methods for predicting the dynamic behaviour of FPV systems installed on irrigation ponds, with a focus on mooring tension and platform movements. Using a dataset derived from numerical modeling trained with historical wind and water depth data, two ML algorithms, the random forest and XGBoost, were employed to develop predictive models. The results demonstrate that XGBoost excels in predicting tension metrics and displacements, whereas the random forest algorithm better estimates yaw rotations. This study also highlights the importance of comprehensive training datasets to capture nonlinear behaviours influenced by the dual mechanical properties of mooring components. Compared with traditional numerical simulations, ML-based predictions significantly reduce the computational time and are a promising avenue for efficient FPV maintenance planning. These findings advance the integration of ML in renewable energy technologies, increasing the sustainability of FPV systems.