Posters

Abstract

Floating offshore wind is a key technology for harvesting wind resources in deep waters, where bottom-fixed foundations become technically challenging and often economically unattractive. Compared to bottom-fixed turbines, floating systems introduce additional fatigue load drivers: wave-excited platform motions couple with aerodynamics and control actions, and this coupling can significantly affect Damage Equivalent Loads (DEL). Consequently, wind farm control strategies that are purely power-oriented (e.g., wake steering through yaw misalignment) may unintentionally accelerate damage accumulation, especially for downstream turbines operating under elevated turbulence intensity and partial-wake conditions. In the literature, life-aware wind farm operation has been explored by embedding damage or lifetime indicators into farm-level control and optimization, typically combining engineering wake models with fast damage evaluation (e.g., surrogate models and/or precomputed tables) [1]. In parallel, computationally efficient fatigue models have been proposed to make damage-aware wind farm optimization tractable. A representative example is the methodology presented in “A model to calculate fatigue damage caused by partial waking during wind farm optimization”, where fatigue damage due to partial waking is estimated through reduced-order wake/turbulence descriptions and load/damage surrogates, enabling direct inclusion of damage metrics within optimization problems [2]. However, these approaches are largely developed and demonstrated for bottom-fixed turbines and do not explicitly capture the wave-induced dynamics intrinsic to floating platforms [1,2]. In the floating context, applying the same methodology in a straightforward way would become computationally prohibitive: to account for the impact of sea states on DEL, one must span a high-dimensional space that includes wind conditions (wind speed, direction variability, wake state, turbulence), sea states (wave height, peak period, wave direction), and control variables (yaw schedules). If tackled with high-fidelity farm simulations (e.g., FAST.Farm) across this combined space, the number of required simulations would be excessive, undermining the practicality of fatigue-aware closed-loop farm control. This work investigates a methodology that drastically reduces the number of high-fidelity simulations required while maintaining reliable fatigue-damage predictions for control-oriented applications. The proposed workflow is organized in two stages and leverages data-driven surrogate modeling. First, a large dataset of high-fidelity single-turbine simulations is generated in OpenFAST for a floating wind turbine under high-turbulence inflow conditions, designed to represent both freestream and waked operation. DELs computed from this campaign are used to train a fast feed-forward neural network that maps representative operating and environmental descriptors to DEL estimates. Second, to capture wake interactions and farm-level effects not fully represented by isolated-turbine training data, a targeted test matrix of four-turbine simulations is performed in FAST.Farm under different wake and wake-steering conditions. These multi-turbine results are used to validate and calibrate the surrogate, improving its reliability when transferred to realistic wind-farm operation. The final outcome is a tool capable of estimating DELs nearly instantaneously from current farm operating conditions, enabling closed-loop yaw optimization that maximizes power while enforcing fatigue-related constraints (or penalties) based on predicted DEL. [1] First experimental results on lifetime-aware wind farm control [2] A model to calculate fatigue damage caused by partial waking during wind farm optimization