the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A global blockage parametrization for engineering wake models
Abstract. Whereas engineering wake models can be used to efficiently provide energy production estimates for wind turbine sites, recent studies indicate the importance of a global blockage effect becomes manifest for larger assets. This global blockage effect is caused by site-scale interactions with the atmospheric boundary layer, and results in a wind speed deficit upstream of the asset. This paper presents an efficient and accurate parametrized global blockage model which integrates into existing engineering wake models. The central idea behind this global blockage model is to interpret the wind farm site as a parametrized porous object, subjected to an ambient flow field. We calibrate and benchmark our model through high-resolution LES model data for a representative offshore site using a calibrated wake deficit shape parameter. Results show significant improvements in turbine-level energy production prediction accuracy when compared to results obtained without any blockage model and results obtained with the local self-similar blockage model. The parametrized global blockage model has a significantly lower computational footprint compared to local blockage models. We conclude that not taking (global) blockage into account sufficiently can yield a tendency to overestimate the strength of the turbine wake deficit effects when calibrating wake deficit shape parameters. Finally, we note that the spatial distribution of (global) blockage and wake deficit errors can easily lead to error cancellation when aggregating over binned wind directions.
Competing interests: All authors are employed at Whiffle Precision Weather Forecasting BV, a limited liability company.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(5023 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 03 Jun 2025)
-
RC1: 'Comment on wes-2025-71', Anonymous Referee #1, 25 May 2025
reply
Please see the attached file.
-
RC2: 'Comment on wes-2025-71', Anonymous Referee #2, 26 May 2025
reply
The paper ”A global blockage parametrization for engineering wake models” by Goedegebure, Folkersma, and Maljaars proposes a parametrized model for global blockage. The model is to be used with engineering wake models and makes the wind speed perturbation due to a wind farm depend on the height of the atmospheric boundary layer. This dependence, which is missing in the most common engineering wake models, has been demonstrated in field experiments and high-fidelity simulations.
The authors use a Gaussian wake model and consider three variations of this model. In the first, the wake model alone is used to simulate the flow. In the second variation, the wake model is coupled to a wind farm blockage model describing a blockage field aggregated from individual turbine induction contributions. The third variation is the proposed new global blockage model. For all three model variations the wake expansion parameter in the wake model is calibrated using LES data for a 90-degree sector of wind directions for a base case wind farm layout. Each model variation is then tested on LES data for all wind directions for the base case layout and for the same 90-degree sector of wind directions for two alternative layouts as well as for the base case layout but considering all wind directions.
The authors make a point of the local blockage model being much slower than the wake model either alone or combined with the new global blockage model. But this hinges on using a default setting in PyWake for the convergence of the effective wind speed. Specifically, that the effective wind speed is changing by less than 1e-6 between successive iterations. This is a ridiculously small change given the accuracy one can presume of the model. I realize that this may be the default setting of the program, but it is hardly necessary. I miss a discussion of why the local blockage model may be inherently slower from a computational point of view than the wake model and the new global blockage model.
The coupling of the global blockage model utilizes some named features and methods in PyWake. The authors should rewrite these parts of the manuscript to make the methodology easier to follow for those reader not familiar with PyWake. Write down the appropriate equations explaining the coupling procedure and then potentially include (maybe in an appendix) a description of how this can be accomplished in PyWake using already existing functionality.
The new global blockage model depends on the height of the boundary layer H, the size of the wind farm (l) perpendicular to the flow direction, and a drag coefficient C_d. The model is built form analogy with blockage in a wind tunnel. The parameter C_d should depend on the geometry (porosity) of the wind farm. The authors choose a value of C_d=1 for simplicity. The wind farm size l is based solely on the front row. These choices may work OK for the regular layouts considered in the paper, but it is hard to trust that they will work equally for all types of layouts. The authors would do well to at least present an avenue for considering the details of the wind farm layout in future work.
The key parameter of the new model is the boundary layer height. This is known to have a significant influence on the wind farm flow. The authors use the mean value of H for four 90-degree wind direction sectors. Given that the LES data on which the work is based are inherently time series this seems like a peculiar choice. Why not run the calculations in a time series manner and fully leverage the dynamics of the boundary layer height? The authors make a point out the differences between performing a summation over directions before or after the errors are calculated. But for the important parameter H they are contend with averaging over large wind direction sectors.
The results are interesting, but without more thoughts on the parametrization of the global blockage model (in particular C_d) and more granularity in the applied boundary layer height it is not clear if the new model is actually useful. The authors should extend the analysis and rewrite the paper to include these items. Another nice addition would be to consider a less regular layout with a different wind rose to assess how well the calibrated model can be applied at a very different site. Then I think the paper would be suitably for publication in Wind Energy Science.
- What is the hub height and rotor diameter of the turbines? This information is needed if one is the replicate the results.
- How is the boundary layer height calculated from the LES data? Specifically, given the many different definitions of the boundary layer height what is the chosen definition in this work meant to signify physically and how is this related to the proposed global blockage model?
- How is the coupling between the wake model and the new global blockage model done? Is it only one-way or is it iterated?
- What was the period for the LES data?
- Was any filtering applied on wind speeds or did all the cases include all wind speeds from the LES?
- From Figure 2 it seems that the chosen local blockage model does not modify the flow field does not modify the flow field downstream of the rotor. However, in reality the flow is accelerated around the rotor. There are other local blockage models that include this effect.
- Line 206: should northeastern wind speed be northeastern wind directions?
- Figure 8: the grey dots are hard to see in the plot. Choose a more distinctive colour.
- Line 253: are predictions errors the RMSE?
- Line 3001: what does deficit inducing wake shape mean?
- Figure 11: the crescent features of the plots, see first panel near the point (14, 14), are so prominent that they call for an explanation. What are we seeing here?
- Figure 11: the caption states that values for the A*case appear more correlated because they are aggregated over all wind directions. It would be useful then to see the results from this case for isolated wind direction (sectors). Perhaps the four quarters of the unit circle.
- Figure 11: I assume there is one dot per turbine, but this should be made clear in the caption.
Citation: https://doi.org/10.5194/wes-2025-71-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
141 | 24 | 7 | 172 | 6 | 3 |
- HTML: 141
- PDF: 24
- XML: 7
- Total: 172
- BibTeX: 6
- EndNote: 3
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1