the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Apr 2025
Wind turbine wake dynamics subjected to atmospheric gravity waves: A measurement-driven large-eddy simulation study
Abstract. Atmospheric gravity waves (AGWs) are large-scale wave-like flow structures commonly generated when atmospheric flows are vertically displaced by topography. These transient phenomena can significantly affect wind turbine outputs and loads; however, their influence on wake dynamics remains poorly understood, posing challenges for accurate wind farm modeling. In this study, we perform large-eddy simulation of wind turbines operating under an atmospheric condition reconstructed by assimilating lidar measurements of AGWs. Our results show that: (i) Low-frequency wake meandering becomes more pronounced owing to large-scale AGW flow structures and intensified smaller-scale turbulent structures. This enhanced meandering, combined with stronger turbulent mixing, accelerates mean wake recovery. (ii) Turbulence kinetic energy (TKE) spectrum in the wake region exhibits a peak Strouhal number of approximately 0.3, although the inflow spectrum peaks at significantly lower frequencies. This observation indicates that, under AGW conditions, wake turbulence generation follows a convective instability mechanism. Notably, faster wake recovery reduces wake shear, leading to lower amplification of TKE. Power analysis for three turbines arranged in a streamwise column further highlights the dominant role of convective instabilities. Large-amplitude, low-frequency power fluctuations observed at the most upstream turbine are significantly attenuated for downstream turbines, as low-frequency velocity fluctuations shift to higher frequencies in the far-wake regions. These findings add further insights into wake meandering and turbulence generation, offering guidance for modeling wind turbine and farm flows under non-stationary atmospheric conditions.
- Preprint
(4438 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on wes-2025-52', Anonymous Referee #1, 07 May 2025
-
RC2: 'Comment on wes-2025-52', Anonymous Referee #2, 15 Jul 2025
WES-2025-52 Review
The paper presents an LES study of a 5 MW wind turbine subject to an atmospheric gravity wave (AGW) simulated using an indirect profile assimilation method such that the inflow velocity profile matches LIDAR measurements at a point at the site of the AWAKEN project, and subject to a non-AGW logarithmic boundary layer profile. For the two onset flow fields several wake characteristics are compared including amplitude of wake meandering, average turbulence along the wake centreline, TKE spectrum at 4D downstream of rotor plane, and recovery of wake velocity. The final section reports the influence of the AGW flow on the power performance of three in-line turbines highlighting that large variations of power due to the AGW experienced by the first turbine are reduced significantly by the second turbine and nearly disappear for the third turbine.
Whilst there are some interesting aspects, several parts would benefit greatly from clarification or more detailed analysis to make the scientific significance clearer, to ensure reproducibility of the analysis and ensure consistency between the different sections of analysis.
1. Extent to which the chosen interval represents flow at the AWAKEN site and more generally. Source of data should be given (line 54). The paper identifies use of location A1 and of 08 June 2023 only with the justification that the vertical extent of the gravity wave spans the rotor layer of the wind farm. It would help the reader to show how often these types of condition occur at the AWAKEN site - if only very rarely then is this relevant for design?, if regularly then what range of AGW vertical extents occur? were similar observations (phase lagged) obtained at down-wind sites (e.g. H on Figure 1 map) or does vertical extent differ over the streamwise spacing between turbine rows? More generally atmospheric gravity waves can occur at other sites and locations so some discussion on how the conditions represent AGW conditions at wind farm sites more generally (e.g. in terms of ABL thickness and AGW wavelength - mentioned to be 2 km in LES ( line 176), compared to 2.5-3 km measured (line 58)? - and amplitude, not only site roughness which is not the only important factor between sites).
2. Extent to which the indirect profile assimilation method reproduces the LIDAR measurements of the selected AGW event. The three frames shown on left hand side of Figure 3 compare LIDAR measurements of time varying onset velocity to the simulated conditions. Lines 94-96 comment that “the present LES not only captures the low-frequency wind speed oscillations by the AGW event but also resolves turbulence structures with higher spatio-temporal resolution’. Whilst the simulation seems to capture the period of the selected AGW event there seem to be other differences that are not mentioned; for example the LES shows larger maximum velocity, possibly larger minimum velocity, change of turbulence over 0.5 < height < 1 during 06:00 to 07:00 (approx). These need to be critically assessed, particularly in the context of line 39-40 “it is unclear whether LES driven by field measurements can accurately capture transient atmospheric phenomena like AGWs”. A quantitative comparison is needed of the measured and simulated conditions. Comparison of profiles - of velocity, turbulence, potential temperature - at specific time-steps during the AGW event would provide greater clarity. There should also be discussion on whether these AGW event predictions can be considered to be mesh independent and sub-grid model independent.
3. Choice of non-AGW conditions used for comparison. The same onset flow profile plots would also be useful to show the non-AGW conditions modelled rather than relying on the comments regarding similarity on lines 116-119 only. Figure 6 shows that ambient TKE is very different between the two cases so is it meaningful to compare wakes in such different turbulence conditions?.
4. Choice of turbine modelled. Please summarise the differences and similarities between the deployed GE 2.8 MW turbine and the NREL 5 MW reference turbine to explain why this substitution was made and highlight the implications of any differences of diameter, hub-height and operating characteristics.
5. Turbine modelling approach. Reference needed for statement on line 104-105 re model choice previously demonstrating good agreement. Since the focus of this study is on locations up to 8D downstream (and for Figures 8 and 12 at 4D downstream) please clarify that the previous demonstration of good agreement for far-wake predictions relates to comparable distances.
6. Meandering results. The analysis focuses on the streamwise increase of amplitude of meandering of the wake center. To relate the observed variations to the two mechanisms identified (lines 143-145) it would be helpful to show that this meandering of wake center is occurring at the AGW period (~10 mins as Fig 3?), and to show the turbulence length-scales (which are not currently stated in the manuscript), or corresponding time-scales, for each case.
7. TKE spectra analysis. This is interesting, particularly the peak sustained at Strouhal Number ~ 0.3. However, the lack of peak at St ~ 0.05 in the wake in AGW case seems to indicate that there are not variations in the wake at the AGW period; does this affect line 141-142? Could the same type of spectra be shown for 8D also to better understand whether the same spectral content persists as the amplitude of meandering increases into the far wake? Is there any explanation available for the higher harmonics observed in AGW case?
8. Wake velocity recovery. As noted on line 199 the distance to the maximum velocity deficit differs between the AGW and non-AGW cases. It is not however shown that the wake form is Gaussian at this point. The points on ‘faster recovery of the mean wake of the AGW’ should relate to the distances after this near-wake region. Over that range the AGW wake recovery is faster but it would be useful to bring out this rate of recovery of velocity more clearly. Differences of wake recovery are attributed to two mechanisms: i) stronger wake meandering due to larger-scale turbulent structures, and ii) higher value of TKE in the AGW case. The earlier sections should be modified to support these statements quantitatively including clarification of: the scales of turbulent structures in the onset flows, that the wake meandering is at the AGW periods, the value of TKE of the non-AGW onset flow.
9. Power attenuation is interesting to observe. However, this spacing is much closer than at the site or for any practical operating conditions so justification is required of the relevance of this layout. Some discussion is also needed on how these observed variations of power relate to the choice of turbine. At this spacing mean power on second turbine is about one-third of that on the leading turbine (hub height velocity is ~ 0.7U0 as figure 8) so fluctuating load relative to the mean seems to increase with downstream rotor position (even though wake TKE similar to ambient TKE). Is the third turbine still within the operating range of the turbine if Uhub ~ 0.7 * 0.7 U0 ~ 2.5 m/s?
Several of the above comments affect statements currently made in the conclusions - particularly regarding turbulence characteristics (length-scales) of the onset flows and the periods affecting wake meandering and wake recovery. The findings regarding power attenuation of downstream turbines should be related to the choice of turbine and the relatively low onset wind speeds. Paragraph of line 239 should be rephrased to focus on the scope of this study, and identify any further developments or evaluations that would be needed to establish robustness for a wider range of AGW conditions and turbine wake studies.
Citation: https://doi.org/10.5194/wes-2025-52-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 190 | 61 | 16 | 267 | 14 | 33 |
- HTML: 190
- PDF: 61
- XML: 16
- Total: 267
- BibTeX: 14
- EndNote: 33
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1