Harmonic linearisation of aerodynamic loads in a frequency‐domain model of a floating wind turbine

• Published in: Wind energy (Chichester, England) England), 2021-08, Vol.24 (8), p.833-856
• Main Authors: Lupton, Richard C., Langley, Robin S.
• Format: Article
• Language: English
• Subjects: Accuracy; Aerodynamic loads; Aerodynamics; Aeroelasticity; Computer applications; Control systems; Design optimization; equivalent linearisation; Flexible structures; floating wind turbine; Frequency analysis; frequency‐domain modelling; harmonic linearisation; Harmonics; Linearization; nonlinearity; Perturbation; Turbines; Wind power; Wind turbines
• ISSN: 1095-4244; 1099-1824
• DOI: 10.1002/we.2605

While detailed aero‐servo‐hydro‐elastic simulation codes for modelling floating wind turbines (FWTs) are available, they achieve high accuracy at the expense of calculation speed. For conceptual design and optimisation, fast solutions are needed, and equivalent linearisation techniques combined with frequency‐domain analysis offers to capture the complex behaviour of FWTs in fast, approximate models. The main aim of this paper is to apply a harmonic linearisation approach to model the aerodynamic loading within a complete coupled model of a FWT, quantifying its performance, and where accuracy is unsatisfactory, to give insight into the causes. Two linearised models are derived from a coupled nonlinear aero‐hydro‐servo‐elastic model, using the OC3‐Hywind FWT as a test case: the typical tangent linearisation derived by numerical perturbation of the nonlinear model and the harmonic linearisation yielding improved representation of the aerodynamic loads. Comparisons against nonlinear time‐domain simulations from Bladed show that it is possible to create a frequency‐domain model of a FWT, including a flexible structure, aeroelastic rotor loads and the effect of the control system, with reasonable accuracy. The biggest source of errors is the presence of additional harmonics caused by nonlinear interactions between loads at different frequencies, rather than inaccurate linearisation of the magnitudes of forces. The computational cost of the harmonic linearisation implemented varies, but in most cases is ∼10× slower than the tangent linearisation and ∼100× faster than the time domain solution.