WEC's co-design optimization

Wave energy presents a substantial, yet underutilized, source of renewable energy characterized by high power density and predictability. This makes wave energy converters (WECs) an attractive option for enhancing the reliability and efficiency of renewable energy systems. Despite their potential, the design and optimization of WECs are fraught with challenges due to the complex marine environment, high costs, time constraints, and significant computational demands inherent in traditional design methodologies. To tackle these challenges, this thesis employs a co-design framework that integrates the physical design of WECs with their control systems. This integrated approach leverages the interplay between the WEC's structural dynamics and control strategies to enhance overall performance, thus addressing the limitations of isolated design processes. However, the comprehensive modeling required for co-design introduces additional computational complexity, particularly in the simulation of WEC behaviors across various design parameters and wave conditions. To mitigate these computational expenses while maintaining high accuracy, this research develops a surrogate model using deep kernel learning (DKL). The surrogate model is designed to predict the transfer function of WECs, which describes the relationship between the wave-induced forces and the resultant motion of the WEC. By training it on data from high-fidelity simulations, it can quickly and accurately predict WEC responses under diverse conditions, significantly reducing the need for expensive and time-consuming simulations. The surrogate model adheres to essential physical constraints, such as symmetry, passivity, and stability, ensuring that the predicted behaviors are realistic and reliable. This methodological innovation allows for efficient optimization within the co-design framework, facilitating the development of more robust and efficient WECs. A case study on a spherical WEC demonstrates the effectiveness of DKL in achieving high prediction accuracy and substantial reductions in computational costs. The model's performance underscores its potential to enhance the design and optimization processes for WECs, making significant strides towards more viable and efficient wave energy solutions. This thesis advances the field of WEC design by presenting a scalable and efficient framework that integrates advanced machine learning techniques with traditional engineering principles. The proposed approach not only addresses the current limitations in WEC optimization but also sets the stage for future research to extend these methods to other renewable energy technologies, promoting innovation and effectiveness in the renewable energy sector.