Data-driven prediction of wave response for a modular floating solar array

Abstract

Floating photovoltaic (FPV) systems in coastal and nearshore regions are subjected to complex wave loads that significantly influence their hydrodynamic performance. In the future, industrial scale FPV projects can consist of numerous floating bodies that are too large to be modeled using existing modelling methods such as Computational Fluid Dynamics. Therefore, there is a need to develop data-driven rapid prediction approaches. This study presents a data-driven framework to predict the heave and pitch response amplitude operators (RAOs) of FPV arrays under different wave conditions. A verified simulation model is developed, generating a dataset that incorporates the key influencing factors, including incident wave angle, wavelength-to-floater-dimension ratio, and mooring type. Random Forest (RF) and Multilayer Perceptron (MLP) models are trained and optimized through grid search cross-validation, demonstrating that both models accurately capture the spatial distribution and magnitude of RAOs, with the MLP model showing superior generalization capability. Interpretability analysis further reveals that the wavelength-to-floater-width ratio is the dominant factor driving RAO responses, while appropriate mooring strategies can effectively suppress motions under extreme conditions. Compared with conventional hydrodynamic simulations, the proposed approach significantly reduces computational cost and enables rapid evaluation of FPV dynamic responses, which provides a potentially workable approach to facilitate various purposes of large-scale ocean solar projects, such as design, monitoring, and digital twins.