Description
Floating Offshore Wind Turbines (FOWTs) currently represent one of the major opportunities in wind engineering as the available shallow water is limited. However, FOWTs are challenging compared to fixed-bottom wind turbines. The floating motion of the platform makes the aerodynamics more unsteady, impacting production, lifetime, and optimal wind farm layout. The resulting system is inherently unsteady, nonlinear, and high dimensional, bringing about significant modelling challenges, associated with a high computational cost. Data-driven modelling has emerged as a promising alternative to first-principle models, reducing the computational time while maintaining sufficient accuracy. A critical first step towards full-scale models is to evaluate the potential of data-driven techniques on systems that are simpler but reflect the essential modelling challenges: nonlinearity, unsteadiness and high dimensionality.Building on prior work on a pitching aerofoil and an oscillating cylinder, this study introduces a new unsteady dataset based on a pitching actuator disk that mimics the aerodynamics at the turbine level of a FOWT. Actuator-disk models are widely used, relatively low-complexity models of wind turbine aerodynamics. Multiple model classes are tested and compared to identify a dynamic model linking the pitching angle to the high-dimensional wake behind the disk. These model classes include Sparse identification of nonlinear dynamical systems (SINDy), Long short-term memory (LSTM) neural network, and state-spaces neural network (SS-NN) approaches such as deep subspace encoders. A set of carefully selected pitching swept-sine motions of different amplitudes is employed to replicate various platform motions.
This work aims to construct a dataset of a pitching actuator disk and evaluate the capability of data-driven methods to produce stable, accurate multi-step-ahead predictions, supporting integration into model predictive control for FOWTs.
| Period | 29 May 2025 |
|---|---|
| Event title | Ai & Fluids 2025 |
| Event type | Conference |
| Location | Chania, GreeceShow on map |
| Degree of Recognition | International |