Hybrid frequency–time domain surrogate-augmented model of dynamics of floating offshore wind turbines

Abstract

Floating offshore wind turbines can harness the more consistent and powerful winds in deep-water offshore areas. However, they face challenges due to complex loading mechanisms across various load cases. Fully coupled nonlinear time-domain response and load analyses are computationally expensive, making them unsuitable for highly iterative design optimisation under all limit states’ constraints. This research presents a hybrid domain (frequency and time) and hybrid paradigm (physics-based, surrogate-augmented) approach that strikes a balance between accuracy and computational efficiency. The motion response is obtained in the frequency domain, while structural loads are computed from the inverse fast Fourier transform of the motion spectra. This approach combines the efficiency of the frequency-domain approach with the ability to model some load nonlinearities in the time domain. To further enhance efficiency, two dedicated surrogate models have been developed and embedded into the workflow. The first surrogate predicts hydrodynamic potential coefficients over the relevant range of wave frequencies, thereby replacing the need for repeated potential-flow computations in the design loop. The second surrogate emulates the time-domain mooring line loads as a function of mooring design and fairlead motion, acting as a fast proxy for more computationally demanding dynamic mooring analyses. The mooring surrogate is agnostic to platform design and environmental conditions, offering broader applicability than existing approaches. Both surrogates are constructed as data-driven, multi-output regression models trained on large datasets generated from higher-fidelity numerical simulations, tailored to provide accurate predictions in milliseconds. This novel approach provides accurate design trend predictions while significantly reducing solution time. It enables fast and comprehensive FOWT design optimisation, accelerating early-stage development and allowing greater complexity from the outset.