Data-Driven Koopman Operator Modeling for Real-Time Nonlinear Dynamics Identification of Floating Offshore Wind Turbines

Real-time structural health monitoring (SHM) of Floating Offshore Wind Turbines (FOWTs) is often limited by the inability of linear models to capture stochastic hydro-elastic coupling and viscous effects. To address this, we propose a data-driven Reduced Order Model (ROM) based on Koopman Operator Theory, which identifies a linear embedding for the underlying nonlinear dynamics. Unlike traditional system identification methods, this framework does not require governing equations; instead, it leverages a physics-informed observable subspace, comprising vibration measurements, to approximate the infinite-dimensional Koopman operator. The approach is validated using the NREL 5MW turbine on an OC3-Hywind spar platform, simulated under wind and wave load conditions via OpenFAST. The model will investigate the strength of dynamic mode decomposition alternatives on isolating coupled platform-tower modes and forecasts structural response with high fidelity. By significantly reducing the computational overhead compared to multi-physics aero-hydro-servo-elastic simulations, this method enables real-time fatigue assessment and digital twinning for offshore assets.