Session: 09-02-06: Wind and Wave Energy: Moorings

Submission Number: 157636

Machine Learning-Based Design Optimisation of a Wec Mooring System 

The design of Wave Energy Converter (WEC) mooring systems involves navigating complex, multi-parametric spaces to achieve commercially viable, reliable, and high-performing solutions. Traditional, iterative methods using time-intensive dynamic simulations are prone to: (i) missing optimal regions of the design space, (ii) revealing unacceptable load cases during the detailed design phase when the number of load cases is expanded beyond the initial set used for screening. This study proposes a faster and more effective design methodology, using a machine learning surrogate model based on recurrent neural networks, which replicates the dynamic behavior of full time-domain OrcaFlex simulations with computational speeds approximately 100,000 times faster. The recurrent neural networks were trained to be wave-seed agnostic, meaning they can fully replicate the underlying physics of the Orcaflex model, and predict loads and motions for any unseen irregular wave train. This enabled extensive exploration of the design space for all required load cases, allowing Pareto optimization to minimize key objectives such as bending moment, vertical loading at the pile, surge, and pitch. The neural network-based surrogate model achieved high accuracy in replicating dynamic responses and identifying optimal mooring architectures, achieving less than 1% mean error across unseen test cases, and less than 10% maximum error. By significantly accelerating the design process and enabling robust optimization across multiple objectives, this machine learning based methodology represents a transformative advancement for the development of efficient, cost-effective, and resilient WEC designs. 

Presenting Author: Oscar Festa University of Southampton

Presenting Author Biography: Postdoctoral researcher from the University of Southampton, researching machine learning methods for time domain modelling and design optimisation of offshore structures.