Assessing engineering wake models against operational data: insights from the Lillgrund wind farm wake steering campaign

TL;DR

This study validates steady-state analytical wake models against synchronous SCADA-LiDAR data from the Lillgrund wind farm, including baseline operation and wake-steering campaigns. Using four LongSim model combinations that vary in deficit, turbulence, wake superposition, and deflection, the authors demonstrate that velocity-deficit predictions exhibit MAEs of ~7–15%, while turbine- and farm-level power errors can reach up to ~23% and ~30%, respectively; some configurations approach LES accuracy at much lower computational cost. The work shows that, despite inflow heterogeneity and near-wake complexities, the analytical models reliably reproduce wake trends and wake deflection under yaw misalignment, supporting their practical use for wake-steering studies. It also highlights limitations such as homogeneous inflow assumptions, blockage effects, and dynamic inflow considerations, pointing to future work on dynamic inflow, refined near-wake modelling, and inclusion of blockage to improve predictive reliability.

Abstract

Validating engineering wake models under real-world operational conditions is essential for improving wind farm performance predictions. This study uses a unique dataset from the Lillgrund offshore wind farm collected during the Horizon 2020 TotalControl project, combining synchronous SCADA and LiDAR measurements under baseline operation (no intentional yaw offset) and active wake steering. Four analytical wake model combinations are assessed, employing different formulations for velocity deficit, added turbulence, wake superposition, and deflection, implemented in the LongSim software developed by DNV. The analysis focuses on time-averaged wake velocity deficits and turbine- and farm-level power output, with model accuracy quantified using mean absolute error metrics. The models reproduce general wake deficit trends and wake deflection across a range of atmospheric conditions, with normalised velocity deficit errors between 7% and 15%. Power prediction errors increase with farm depth, with turbine-level errors between 3% and 23% and farm-level errors between -13% and +30%. Some analytical models achieve accuracy comparable to reported LES results while requiring substantially lower computational cost. The results highlight the value of field campaigns for benchmarking engineering wake models and inform trade-offs between model fidelity and operational practicality for wake steering applications.

Figures (17)

• Figure 1: (a) Top-down view of the Lillgrund wind farm, showing the installed LiDAR scanners, relative to turbine D08 position. The LiDARs on turbines A07 and C07 measured the velocity within the wake transect synchronously, as represented by the black crosses and detailed in (b). The LiDAR on turbine B08 measured the inflow velocity, with its centreline indicated by the blue line and detailed in (c), alongside its side view in (d), where the square represents the hub-height measurement of horizontal velocity $U_h$ and its direction $\theta_h$. Adapted from Sood et al.sood2022comparison
• Figure 2: (a) Lillgrund data timeline in 10-minute timestamps. (b) Wind rose depicting wind speed and direction at hub height, based on the inflow measurement LiDAR data during synchronous timestamps.
• Figure 3: (a) The ratio of turbulence intensity (TI) measured with the met-mast to the turbine D08 nacelle's wind speed, grouped by wind speed bins of 1m/s, as reported by Göçmen & Giebel goccmen2016estimation. (b) TI from the synchronous timestamps, filtered for $4\textrm{m/s}<U_h<15\textrm{m/s}$ and $135^\circ<\theta_h<320^\circ$, measured using the inflow LiDAR and SCADA from the D08 nacelle wind speed; the black dots represent D08 nacelle TI multiplied by the ratio.
• Figure 4: (a) Averaged pitch angle of turbine D08 from the filtered synchronous timestamps. (b) Thrust coefficient curves of modelled non-yawed turbines with representative pitch angles. (c) Normalised power curves compared against D08 field data. The legends in (a) and (b) also apply to subfigure (c).
• Figure 5: Baseline cases (from $\textrm{a}_0$ to $\textrm{g}_0$) inflow conditions: (a) Normalised horizontal velocity profile along its range gate, where the dashed lines represent the power-law fitting. (b) Normalised wind veer profile along its range gate, where the dashed lines represent its linear fitting. The framed legend in (b) applies to both subfigures.