the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Aug 2025
Understanding Cluster Wake-Induced Energy Losses off the U.S. East Coast
Abstract. This study seeks to advance our understanding of energy losses caused by wind farm cluster wakes off the U.S. East Coast by utilizing advanced numerical models in conjunction with real-world, available data on existing and planned offshore wind sites. To this end, we have run simulations of existing and planned U.S. offshore wind lease areas using a typical-meteorological-year approach with a GPU-based Weather Research and Forecasting (WRF) model, where lease area layouts are generated based on most up-to-date project capacity information for each individual lease areas. To evaluate wake losses, we use an energy-loss-based definition of "wake shadow", as opposed to the traditional wind speed deficit assessment. A key insight from this study is that large wind speed deficits do not necessarily translate into significant energy losses. In addition, our results indicate that the conventional wind speed deficit method may underestimate the size of the wake area by up to 30 % compared to the proposed energy loss approach. These findings highlight the need to consider both wind speed deficits and energy losses when evaluating the wake effects of offshore wind farms and assessing future offshore wind development.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Wind Energy Science. Furthermore, Mike Optis is the founder and president of Veer Renewables, a for-profit consulting company that uses a wind modeling product, WakeMap, which is based on a similar numerical weather prediction modeling framework as the mesoscale simulations described in this paper.
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 paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(15445 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 20 Sep 2025)
-
RC1: 'Comment on wes-2025-154', Anonymous Referee #1, 21 Aug 2025
reply
General Comments
Overall, this manuscript is well-written and supported by an appropriate number of references, which demonstrates the authors’ thorough engagement with the relevant literature. The use of a GPU-based WRF model is an excellent choice, and the configuration decisions—such as setting the TKE factor to 1 and employing a 1 km mesh resolution at the wind farm—are well-justified and technically sound.
I particularly appreciate the authors’ decision to focus on power loss rather than velocity deficit maps, as this approach provides a more meaningful metric for assessing wind farm performance. The AEP loss map presented in Figure 7 is especially informative and visually compelling.
The analysis showing that large velocity deficits at wind speeds above the turbine’s rated speed (11 m/s) have a limited impact on overall wind farm production is interesting. While this may not represent a groundbreaking finding, it is still valuable to publish, given the historical reliance on velocity deficit maps in WRF parameterization studies. This contribution helps clarify the limitations of traditional approaches and reinforces the importance of using power-based metrics.
Major Revisions
1- One of the most relevant findings in the manuscript relates to stability differences between the surface and the rotor area. This is an important aspect of wake modeling, and I encourage the authors to expand this analysis. Specifically, could you include a comparison with 1/L at the surface, which is a standard WRF output and widely used in the literature? Additionally, it would be helpful to investigate whether there is a larger mismatch between Ri at the surface and at rotor height near the shoreline. Please also clarify in the text how stability classes were defined using Ri values, as this is currently only indicated in Figure 4 and Table 3.
2- The configuration of WRF and the chosen parameterization is appropriate and well-executed. However, the study currently relies on a single model and configuration, which limits its robustness and generalizability. To strengthen the work, I recommend including a comparison with another widely used industry model, such as the Turbopark wake model (for at least one of the wind farms) or the Volker parameterization. This would provide additional context and enhance the credibility of the conclusions.
3- Regarding the contour maps presented in Figure 8, which are among the most impactful results for decision-making on wind farm siting: Is it necessary to simulate a full year to generate these maps? A sensitivity analysis using a shorter simulation period would be highly valuable. Furthermore, if a new wind farm were to be added, would it be necessary to rerun the entire cluster simulation, or could the new farm’s results be combined with the existing data? Addressing these questions would significantly improve the practical applicability of the study.
Minor Revisions
1- The improved parameterization proposed by Vollmer et al. (2024) could potentially influence the conclusions of this work. Would it be feasible to conduct a limited test to confirm that the main findings remain consistent under this updated approach?
2- Additionally, please elaborate on the rationale for selecting the three variables used in the TMY methodology. Would including additional variables such as TKE or stability indicators (Ri or 1/L) improve the representativeness of the dataset? Including a standard wind rose plot for the TMY dataset would also enhance the clarity of the analysis.
3- It would be useful to verify whether a single turbine in the domain reproduces the input power curve when using the parameterization, particularly under stable conditions where low-level jets affect the rotor area. This would help clarify whether some of the observed losses are due to wake effects or to the parameterization’s response to vertical wind profiles.
4- In Figure 6, consider presenting energy loss aggregated at the wind farm level, in addition to the 1 km grid cell resolution. Furthermore, could you explain the energy loss observed far from the wind farms in this figure? Is this related to the well-known WRF parameterization numerical errors in stormy conditions?
5- Finally, for context, are velocity deficit and power loss maps actually used by government agencies for planning purposes? If so, please provide a reference where this application is documented, as this would strengthen the practical relevance of the study.Citation: https://doi.org/10.5194/wes-2025-154-RC1 -
RC2: 'Comment on wes-2025-154', Anonymous Referee #2, 02 Sep 2025
reply
The manuscript uses WRF to study the losses caused by cluster wakes off the US East coast. They simulate a reconstructed typical meteorological year and provide a framework for layout generation of future lease areas. They then evaluate the result using different metrics, specifically wind speed related, or energy (power) related which means the wind speed signal is evaluated with the power curve.
This work addresses the currently very relevant cluster wakes topic. However, the content better fits to a technical report which could be used in for instance policy making. In my opinion, there is hardly any scientific relevance or novelty to this work, see specific comments below. For this reason, I regretfully recommend the manuscript in its current form to be rejected.
Specific comments
- The study compares wind speed deficit with energy losses, with the main conclusion: “A key insight from this study is that large wind speed deficits do not necessarily translate into significant energy losses.” (l. 6-7). This is directly linked to the fact that above rated wind speed, wind speed deficits will not affect the turbine’s generated power.
I fail to see the novelty of this observation, as this has been known for as long as turbines have operated with a constant rated power. That this is now applied to cluster wakes rather than individual turbine wakes, is in my opinion insufficient to justify a publication. - The study also investigates the effect of stability on the wake losses, leading to the conclusion: “… the greatest energy loss occurs when wind speeds are below-rated with stable stratification.” (l. 372). Also this result follows existing literature, showing that cluster wakes recover more slowly under stable conditions.
- The analysis of the wake area in Section 4.2 mostly presents conclusions that are fully expected from the analysis in Section 4.1, i.e. difference in wind speed deficit and energy loss approaches presented in Sections 4.1 and 4.2 are the same. What is more interesting in this section is the relation of F_wakearea to the capacity density.
- The most interesting contribution of this study comes, in my opinion, from Section 4.3 focusing on the use of different ways to calculate the ‘same’ metric. This addresses uncertainty or ambiguity that is easily introduced. My recommendation for a revised manuscript would be to shift to focus to an uncertainty assessment, which can include the use of different metrics (current study of wind speed deficit versus energy loss), different understandings of the same metric (eq. 7-8), but also assumptions about layouts, turbine types and capacity densities that are taken here at the beginning of the study. Additionally, model setup choices (like value of TKE coefficient) should be discussed.
Technical corrections
- 25 and throughout the manuscript: order references chronologically or alphabetically
- Section 2.1: add (numbered) subsections for clarity
- 134: this modification addresses induction, not blockage. Also, the effect of using this (or other) wind farm parameterizations on your results should be discussed.
- 3: recommend to also show the original 15MW curves to see the effect of peak shaving. Also justify why peak shaving is only applied to the 15MW turbine.
- 207: limitation of taking the stability of the entire rotor layer should be noted, as it becomes a different metric. It is for instance not valid any longer for turbulence closure as it does not evaluate for local turbulence generation or suppression. Also note that inhomogeneity has a huge effect on this variable, think marine boundary layers and LLJs within the rotor area.
- 233: I believe the reference should be to Section 4.1 (not 3.1)
- 5: not a new metric, just 1 minus efficiency. This should be noted.
- 271-272: These farms were not introduced. Rather refer to North/Central/South like in the rest of the manuscript.
- 6: justify that it is fair to look at wind speed at one height, while turbines with different hub heights are simulated
- Section 4.1: add (numbered) subsections
- 286: It was unclear that before you were looking at the 5% threshold only.
- Fig 8: explain in text why the results in the North-East deviate from the other wind directions. For all wind direction, energy loss > WS loss, except in the North-East.
- 345 typo: “climate.,”
- 348-350: This is an interesting result from a modeling point of view and deserves more attention. Should be discussed as another source of uncertainty.
- Figure B4: should be moved to Section 2.2.
Citation: https://doi.org/10.5194/wes-2025-154-RC2 - The study compares wind speed deficit with energy losses, with the main conclusion: “A key insight from this study is that large wind speed deficits do not necessarily translate into significant energy losses.” (l. 6-7). This is directly linked to the fact that above rated wind speed, wind speed deficits will not affect the turbine’s generated power.
Data sets
Understanding Cluster Wake-Induced Energy Losses off the U.S. East Coast Geng Xia https://doi.org/10.5281/zenodo.15078171
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 316 | 43 | 7 | 366 | 4 | 3 |
- HTML: 316
- PDF: 43
- XML: 7
- Total: 366
- BibTeX: 4
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1