Seasonal variability of wake impacts on US mid-Atlantic offshore wind plant power production

Rosencrans, David; Lundquist, Julie K.; Optis, Mike; Rybchuk, Alex; Bodini, Nicola; Rossol, Michael

doi:https://doi.org/10.5194/wes-9-555-2024

Articles | Volume 9, issue 3

https://doi.org/10.5194/wes-9-555-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/wes-9-555-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 9, issue 3

Research article

|

14 Mar 2024

Research article |

| 14 Mar 2024

Seasonal variability of wake impacts on US mid-Atlantic offshore wind plant power production

David Rosencrans, Julie K. Lundquist, Mike Optis, Alex Rybchuk, Nicola Bodini, and Michael Rossol

Download

Final revised paper (published on 14 Mar 2024)
Preprint (discussion started on 04 May 2023)

Interactive discussion

Status: closed

RC1:
'Comment on wes-2023-38', Anonymous Referee #1, 24 May 2023

The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-38/wes-2023-38-RC1-supplement.pdf

Citation: https://doi.org/10.5194/wes-2023-38-RC1
- AC1: 'Reply on RC1', David Rosencrans, 18 Sep 2023
  
  The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-38/wes-2023-38-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/wes-2023-38-AC1
CC1:
'Comment on wes-2023-38', Mark Stoelinga, 02 Jun 2023

Thanks to the authors for this very good contribution. I have a few comments and questions that I'd like to share (references by line number or section).
181-182: I think centered RMSE (cRMSE) is essentially the same as what I’ve heard and referred to as bias-corrected RMSE (or BCRMSE), in which you first calculate the mean model bias error, subtract it from all the model values, then calculate RMSE. And, I believe both are essentially equivalent to the standard deviation of the errors as well. All that is neither here nor there. However, I do think the sentence in lines 181-182 should be clarified, to say that “a value of 0 for cRMSE indicates that all values, after removal of the respective model or measured means, lie on the 1:1 regression line”.
189 (paragraph): Might be good to show model versus measured mean shear exponent, a metric that the wind industry uses extensively and is highly familiar with its typical range of values.
325-326: There is an interesting result in Fig. 8 that you do not comment on, which is similar to behavior other have seen and commented on (including, I believe, one or more of you in previous work, and myself). What I’m referring to is the opposite effect of TKE amount in the near-project versus distant wake environment. Within and near the project, behavior is intuitive: higher TKE dissipates wakes and leads to smaller waked wind deficits. However, farther away, as evidenced by the distance northeastward of the first (0.5 m/s) contour, as well as the area of this contour reported in the text, it is actually slightly farther (and covers more area) with TKE than without it. In other words, at distance, higher TKE actually helps wakes, whereas near or within the project it hurts wakes. I saw the same behavior, and I’m certain you and others have commented on it previously. Do you have any new insights into this behavior?
Appendix E. The authors and I have had discussions in the past about the nature of the noise seen in difference fields (turbines minus no turbines wind speeds). I’m not opposed to the idea that they are purely numerical; I agree that is the most likely explanation. However, I still consider it possible that even the distant differences are perhaps partly physical rather than numerical. They tend to occur in an unstable boundary layer or in convective scenarios. These scenarios are characterized by small-scale, high-amplitude, chaotic structures (convective cells) whose initiation locations are random and probably sensitive to even the smallest perturbations, which may include very subtle and fast-moving gravity waves or other disturbance triggered by the presence of the turbines. For the purpose of energy production, though, they are probably inconsequential because they tend to cancel each other out when averaged either spatially or temporally.
I hope these comments and questions help.
Best regards, Mark Stoelinga

Disclaimer: this community comment is written by an individual and does not necessarily reflect the opinion of their employer.

Citation: https://doi.org/10.5194/wes-2023-38-CC1
- CC2: 'Reply on CC1', Mark Stoelinga, 02 Jun 2023
  
  Well, I see that I missed a key paragraph on my first read, in which you addressed the question of higher TKE actually producing longer-lasting wakes at distance. Sorry for missing that key point. The explanation makes sense. -Mark S
  
  Disclaimer: this community comment is written by an individual and does not necessarily reflect the opinion of their employer.
  
  Citation: https://doi.org/10.5194/wes-2023-38-CC2
- AC2: 'Reply on CC1', David Rosencrans, 18 Sep 2023
  
  The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-38/wes-2023-38-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/wes-2023-38-AC2
RC2:
'Comment on wes-2023-38', Anonymous Referee #2, 13 Jun 2023

This study characterizes the wake impact of proposed wind farm in the mid-Atlantic Outer Continental Shelf through a series of model simulations using the Weather Research and Forecasting model. It includes an extensive set of one year long simulations, including a simulation without wind farms, a simulation for Vinyard Wind and a simulation including all lease areas. In addition, two different configurations added TKE are considered (0 % and 100 %). Finally, an additional 4 month long simulation is performed containing also the call areas.
This is an impressive set of simulations with a through analysis for a relevant topic. I find the analysis on the added TKE particularly interesting, especially the conclusion that the impact on power is small on average. I recommend publishing of this manuscript after addressing the numerous minor points below.
Minor comments

- line 68: While 12 MW turbines seem to be similar enough to the 13 MW turbines to be installed at Vineyard Wind, I wonder how realistic the assumption is for the other lease and especially the call areas, since those will be build later than Vineyard Wind. Also how sensitive are your results to the chosen turbine type?
- line 73 - 74: Why did you choose this period and not a regular calendar year? Do you run continuously or restart the model after a certain period?
- line 79 - 81: You don't mention section 3
- line 95 ff: Please also provide the WRF option number in addition to the reference
- Figure 1 caption: last sentence, double mentioning of "red"
- line 107 - 114: How realistic is the assumption of regular layout within the areas? To reduce internal wake effects, the turbines might be better placed in an irregular layout
- Figure 2a: Where is region 1? Either start numbering at 1 or mention region 1 as below cut-in
- Figure 3: It would be nice to relate the wind rose to the "regions" in Figure 2. E.g. green could be capped at 11 (below rated power) and one color could be used for region 3. Also "m/s" should be formatted with negative exponents according to the guidelines
- Line 179: Does removing the periods induce a bias? E.g. are they related to the same period / stability category?
- Line 170 - 183: Why do you choose these metrics? How do they compliment each other?
- Line 189 - 196: Following up on the previous comment, how do you interpret the results that you obtain for the different error metrics? E.g. something along those lines: "the results correlate well in time but have an offset ...". This should also be discussed for the stability based analysis
- Line 203: You do not describe, which metric you use to classify stability. I assume you are using the same that you use in section 2.7. Consider to move section 2.7 before section 2.6 so that the reader doesn't need to guess.
- Section 2.7: You discuss in Appendix B that the Obukhov length only represents the surface characteristics. Why do you stick to this classification? Also Appendix B should be referenced in section 2.7. Have you estimated the sensitivity of your results to this particular metric? Platis et al. (2021) suggest that depending on the stability metric the results can vary quite a lot (Platis, A., Hundhausen, M., Lampert, A. et al. The Role of Atmospheric Stability and Turbulence in Offshore Wind-Farm Wakes in the German Bight. Boundary-Layer Meteorol 182, 441–469 (2022). https://doi.org/10.1007/s10546-021-00668-4)
- Line 249 - 251: This wake length estimation seems to be too simplified: What about wake turning? What about other wind directions? Arguably the wind rose does show predominant winds from south-west, but other wind directions are also present. In those cases the wake length will be underestimated. To understand your method it would help to draw the line in figure 1.
- Line 270: Reference Appendix E
- Line 304 - 305: This sentence is difficult to understand. Please revise.
- Line 311 - 319: These results could be much more neatly presented in a table instead of text form.
- Line 325: "although areal coverage is larger from reduced wind speed replenishment". What do you mean by this?
- Line 326 - 327: According to the numbers that you present for stable stratification the waked area is actually larger for TKE_100 (16404 km²) compared to TKE_0 (16060 km²). This contradicts with your conclusion in this sentence. Please clarify.
- Line 341 - 345: Again a table would facilitate a comparison between scenarios
- Line 349: You reference D1 here, but D1 only shows TKE_100 and thus the differences due to different TKE levels cannot be assessed.
- Figure 9: Sub-figure titles are (a) for all
- Line 361 - 362: Can you provide the power losses averaged over the four month for VW_only and VW_waked for comparison?
- Section 3.3.1: You show also diurnal variations, but these are not discussed. Please add this.
- Line 383 - 398: It seems a bit counter-intuitively that losses are not additive, i.e. internal losses + external losses != total losses. While the proposed loss estimates (9) and (10) do make sense, they do not share the same reference (P_VW_only vs P_NWF), which makes it more difficult to compare.
- Line 402: I understand the energy demand estimates are taken for present day? Are there estimates on how the energy demand will change until CA and LA are build?
- Line 408 - 409: Could you add another line in figure 11 representing the stability conditions. This would make it easier to see that the power production is indeed more closely linked to hub-height wind than stability.
- Line 424: Reference figure 2 here again to remind the reader of the definition of region 2 and 3
- Figure 13 caption: "black dots indicate turbine locations": suggesting to add "in TKE_0 and TKE_100", since in NWF they are not included
- Line 508: It would be interesting to discuss, how the difference due to added TKE amount compares to the difference due to different PBL schemes. You mention Rybchuk et al. (2022) at some places through the paper, but don't compare the effects due to PBL schemes and added TKE amount directly.
- Line 537: What do you mean by "the differences ... are precise"?
- Appendix A: The mixture of discussion on variability due to added TKE amount and the special case during calm winds between 12:00 and 15:20 on 12 July is confusing. These two aspects should be kept separate.
- Line 555: the first sentence is a bit difficult to understand. The difference between TKE_0 and TKE_25 seems to be more than 15 to 20 m
- Line 565: The way you reference figures is sometimes confusing to me. For instance, I would reference Figure B1 here as "stratification at the E05 and E06 (Fig. B1) lidars exhibits similar seasonal variability to Vineyard Wind (Fig. 6)". Since vineyard wind is shown in Fig. 6 and not in Fig. B1. Please also check other parts of the manuscript. Note also that you wrote "E05" twice.
- Figure D1: Colorbar is missing; is the upper row just a zoom of the lower row?
- Figure E1: "at which the map occurs" -> suggestion "of the map"
- Line 643 - 646: Difficult to understand. What do you mean by "poses a threat to power estimations". I don't understand the contrast "although ..., we show noise occurring in the SE ..." and why this "underscores the point that ... should only show differences within the wake". Please clarify
- Line 660: Is there a link missing for "OpenEI_link"?
- Line 715: Missing DOI
- Line 717: Missing URL
- Line 839: Missing URL
- Line 844: Missing DOI

Citation: https://doi.org/10.5194/wes-2023-38-RC2
- AC3: 'Reply on RC2', David Rosencrans, 18 Sep 2023
  
  The comment was uploaded in the form of a supplement: https://wes.copernicus.org/preprints/wes-2023-38/wes-2023-38-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/wes-2023-38-AC3

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by David Rosencrans on behalf of the Authors (18 Sep 2023) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (08 Oct 2023) by Andrea Hahmann

RR by Anonymous Referee #3 (07 Nov 2023)

Suggestions for revision or reasons for rejection

This paper is well-written, generally well designed and deserving of publication. I noticed that there have been previous review rounds, but I was not involved in those. However, I would like to raise some points that, in my opinion, should be addressed to enhance the paper's quality.

Firstly, it would be beneficial to provide more information about the model's grid setup. The distance between the outer edges of the small domain and the outer edges of the large domain is rather small, and I wonder if the lateral nesting zone of the small domain is not in the spatial spin-up zone of the larger domain. Even though the model setup relies on a previous study, it would be good to clarify this.

I hold the opinion that using (bias-corrected) Root Mean Square (RMS) and correlation as metrics for evaluating model performance may not be the most suitable choice. Even though spectral nudging is applied in the outer domain, the flow in the inner domain can evolve freely. A small deviation in a weather system's position or timing could result in a double penalty effect. Therefore, for this type of studies, it is crucial to focus on getting the statistics right, rather than precisely timing weather systems. In my view, bias and quantification of distribution metrics are more relevant for such studies. I recommend a brief discussion of this issue when describing the evaluation metrics.

I would argue that there should be only one definition for external loss. The total loss can be accurately represented by equation 11, as it is defined based on the reference no-wind farm simulation. Similarly, the internal loss for only the Vineyard Wind farm is correctly calculated, as it also involves dividing by the no-wind farm simulation. However, I have reservations about the correctness of equation 9. Your reference here is the power from the Vineyard Wind, rather than the no-wind farm simulation. The definition of external losses should be the difference between the run that includes all wind farms (combines internal and external loss) and the run with only the Vineyard Wind Farm (internal loss), divided by the power from the no-wind farm simulation (which is the reference). This is equivalent with Equation 12.

The averaging of percentual power loss might be incorrect: I would argue that the percentual power losses, is the total power loss over a period considered (could be total year, or the January 0-1UTC periods), divided by the power of the reference situation (no wind farm) over the same period. However you describe that first you calculate the percentage loss over an hourly window and then these percentages are averaged. This is not the same.

In many climatological studies, the Perkin Skill Score is utilized to compare distributions from climate models for various variables. Both the Earth Mover's Distance (EMD) and the Perkin Skill Score are closely related metrics. It would be valuable to include a brief description of how those two compare. Personally, I find the Perkin Skill Score very intuitive because it is dimensionless and assesses the overlap between distributions. I wonder how the EMD, which is expressed in meters per second, depends on the wind speed itself. Do distributions with higher wind speed also have higher EMD even though the overlap is similar? Does this influence the comparison between stable and unstable that do have different wind speeds?

In your analysis about the different stability classes, you mix the effect of the wake with the fact that the wind rose is likely rather different during the different stability classes. I would appreciate a clearer distinction between the effects caused by variations in wake behavior and those resulting from different wind patterns. To analyse this, you have to include wind roses for both unstable and stable conditions. I expect substantial differences, which could be attributed to seasonal variations. To gain deeper insights, stratifying the data according to wind directions might be a valuable addition.

I find the Turbulent Kinetic Energy (TKE) coefficient sensitivity study interesting and informative. However, I would argue that it does not fully represent uncertainty quantification. There are numerous other uncertain parameters, particularly in wind farm parameterization, which require a more comprehensive approach for uncertainty quantification. I suggest to remove the term uncertainty quantification in abstract and conclusions. You can add a discussion of how this would be a first step towards uncertainty quantification in a discussion.

Appendix A, in my view, is very interesting but not yet mature enough for inclusion in the scientific paper. While it is interesting to study the effect of Turbulent Kinetic Energy (TKE) on surface heat and moisture flux, as well as planetary boundary layer heights, it is a different topic that would benefit from additional model evaluation, particularly concerning those variables.

It would be helpful to provide references or an explanation of the procedure for scaling the 12 MW turbine to the 15 MW turbine.

I wonder how the wind direction statistics have been conducted, especially given that wind directions range from 0 to 360 degrees. A brief explanation or clarification in this regard would enhance the reader's understanding.

Regarding Figure 16 and the related description, I am unsure of the significance of the difference between TKE=%100 and TKE=0%. This difference could potentially be attributed to variations in the timing or positioning of weather systems unrelated to the parameterisation. The explanation of Kelvin-Helmholtz instabilities appears speculative and requires a more detailed analysis. I recommend the removal of this analysis from the paper unless more substantial evidence and analysis can be provided.

P20: the maximum average wake wind speed deficit: it is unclear what this refers to: is it the maximum deficit in space? Is this within the wind farm or outside? And why do you divide by the average hub-height wind speed to get a percentage rather than obtain the maximum percentual average wind speed deficit? This is not necessarily the same. Would be good to add some clarifications.

Hide

RR by Anonymous Referee #1 (29 Nov 2023)

ED: Publish subject to minor revisions (review by editor) (13 Dec 2023) by Andrea Hahmann

AR by David Rosencrans on behalf of the Authors (09 Jan 2024) Author's response Author's tracked changes Manuscript

ED: Publish subject to minor revisions (review by editor) (21 Jan 2024) by Andrea Hahmann

Dear authors

I have reviewed the revised version of your manuscript, and I am happy to see that you have implemented most of the two reviewer suggestions. Congratulations on a job well done!

However, I have some further suggestions to help improve the readability of your article and correct some technical problems:

1. Caption Fig. 3: The location of E06 is shown as the red diamond and E05 as the red triangle.
2. L160: Should be ‘TKE advection is turned on.’
3. You use r^2 in Figure 6 but defined CC in Equation 6. I would use r^2 (or /rho) for correlation, as is commonly used.
4. You use a big V for wind in some places (Eqs 6-8) and then a little v in the shear. It would be best to be consistent.
5. L190: “production of turbulence (Eq. 5): “ remove Eq. 5. A number is already on the equation. Also, in Eq 6-8 and 9. and so on.
6. I think long underscores in equations are very ugly and difficult to read. Why not use “O” (or “L”) for observations (lidar) and “M” (“W”) for modeled (or WRF-simulated)? The same applies to TKE (usually k in equations).
7. Many of the figure captions are incomplete:
1. Captions of Figures 5,6, 7, 8, and 9: “at the E05 (blue) and E06 (red) lidar locations.’
2. Time periods are missing in Figs 5, 6
3. It would be easier to compare lidar and WRF simulated shear if the curves were in the same graphic.
8. Again, in Eqs 13-16, could you shorten the names and underscores? Could the equations be condensed into a single one?
9. Figure 17: The difference in power production between TKE_100 and TKE_0 is shown in MW
10. As a final recommendation, please follow the WES style guidelines at https://www.wind-energy-science.net/submission.html. Some of your journal abbreviations need to be corrected (this will save you time later in the article production).

Andrea Hahmann, WES Associate Editor

Hide

AR by David Rosencrans on behalf of the Authors (23 Jan 2024) Author's response Author's tracked changes Manuscript

ED: Publish as is (24 Jan 2024) by Andrea Hahmann

ED: Publish as is (01 Feb 2024) by Carlo L. Bottasso (Chief editor)

AR by David Rosencrans on behalf of the Authors (01 Feb 2024) Manuscript

Short summary

The US offshore wind industry is developing rapidly. Using yearlong simulations of wind plants in the US mid-Atlantic, we assess the impacts of wind turbine wakes. While wakes are the strongest and longest during summertime stably stratified conditions, when New England grid demand peaks, they are predictable and thus manageable. Over a year, wakes reduce power output by over 35 %. Wakes in a wind plant contribute the most to that reduction, while wakes between wind plants play a secondary role.