Abstract
Wave energy converters (WECs) offer a promising source of renewable energy, but they require a careful joint optimization of both their geometry andcontrol strategy in order to cope with the complexity of marine environments. This joint optimization process is hindered by high computational costs. This paper presents an integrated co-design framework that simultaneously optimizes the WEC physical structure and control system using deep kernel learning (DKL) as a surrogate model. The DKL model, trained on high-fidelity simulations, enables efficient and accurate predictions, while maintaining essential physical constraints. A case study on a spherical WEC shows significant improvements in computational efficiency and a suitable prediction accuracy, advancing WEC design methodologies.