Concurrent Probabilistic Control Co-Design and Layout Optimization of Wave Energy Converter Farms using Surrogate Modeling

TL;DR

This paper presents an integrated system-level approach for optimizing wave energy converter farms by combining sizing, control, and layout design using surrogate models to handle complex hydrodynamic interactions efficiently.

Contribution

It introduces a data-driven surrogate modeling framework with neural networks and many-body expansion principles for concurrent optimization of WEC farm design.

Findings

Effective surrogate models for hydrodynamic interactions

Optimized WEC farm configurations with up to 10 devices

Reduced computational complexity for large-scale array design

Abstract

Wave energy converters (WECs) are a promising candidate for meeting the increasing energy demands of today's society. It is known that the sizing and power take-off (PTO) control of WEC devices have a major impact on their performance. In addition, to improve power generation, WECs must be optimally deployed within a farm. While such individual aspects have been investigated for various WECs, potential improvements may be attained by leveraging an integrated, system-level design approach that considers all of these aspects. However, the computational complexity of estimating the hydrodynamic interaction effects significantly increases for large numbers of WECs. In this article, we undertake this challenge by developing data-driven surrogate models using artificial neural networks and the principles of many-body expansion. The effectiveness of this approach is demonstrated by solving a concurrent plant (i.e., sizing), control (i.e., PTO parameters), and layout optimization of heaving cylinder WEC devices. WEC dynamics were modeled in the frequency domain, subject to probabilistic incident waves with farms of $3$, $5$, $7$, and $10$ WECs. The results indicate promising directions toward a practical framework for array design investigations with more tractable computational demands.