Presentations

Abstract

This study, conducted by SENER together with TECNALIA, focuses on the development of a digital twin for the 15 MW IEA wind turbine installed on a HiveWind semi-submersible floating foundation. Utilizing artificial intelligence, the digital twin characterizes both the dynamic behaviour (rotations and translations across three axes) and static and dynamic forces acting on structural elements and mooring lines. The design leverages data from simulations in the OpenFAST environment, enabling rapid iteration and model refinement in the absence of real-world data. Through supervised learning, predictive models for loads on mooring lines and tower/platform dynamics have been created. The core of the model involves a Random Forest (RF) algorithm, which is validated using cross-validation and iterative predictions. Despite strong performance in interpolation, with maximum mean absolute percentage errors around 5-6%, challenges remain in extrapolating data under extreme metocean conditions. To address this, a Multi-Layer Feed Forward (ML-FF) neural network was introduced, showing improved performance.