Synthetic wind scenario generator from on-site SCADA for wind turbine digital twin

Wind simulation is a critical step for wind turbine lifetime assessment: to accurately represent turbine behaviour across its lifespan, we need accurate wind scenarios. In this work, we propose a methodology for generating synthetic full-field turbulent wind scenarios from sparse, high-frequency SCADA (Supervision Control and Data Acquisition) collected across multiple wind farms.

The proposed methodology is a two-stage process. First, a physics-driven stochastic model learns wind data characteristics from low-frequency measurements extracted from high-frequency sparse SCADA. Second, the pipeline generates a low frequency signal that reproduces the observed spectral content, marginal distribution, and autocorrelation.

We build upon this first stage with a high-frequency turbulence generated via *PyConTurb* [1], which implements IEC/Kaimal coherence models to produce spatially coherent 3D velocity fields across the rotor plane. Our tool is first calibrated using learnt parameters from raw high-frequency SCADA data, then blended with the first signal, which is used as a constraint.

The pipeline outputs TurbSim-compatible BTS files [2] , enabling use in SeaHowl [3] aeroelastic simulations. This industry-standard output enables a ready-to-use wind scenario in both simulation pipelines and machine learning analysis tools.

[1] (Rinker, J. M. (2018). PyConTurb: an open-source constrained turbulence generator. _Journal of Physics: Conference Series_, _1037_, 062032. https://doi.org/10.1088/1742-6596/1037/6/062032) 
[2] (Jonkman, B J (2006). TurbSim User's Guide. https://doi.org/10.2172/891594) 
[3] (De Lataillade, T., Yu, W., Pallud, M., & Capaldo, M. (2024). SEAHOWL: Partitioned Multiphysics and Multifidelity Modelling of Wind Turbines with Monolithically Coupled Elastodynamics. _Journal of Physics: Conference Series_, _2767_(5), 052051. https://doi.org/10.1088/1742-6596/2767/5/052051)