the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Enabling Control Co-Design of the Next Generation of Wind Plants
Andrew P. J. Stanley
Christopher J. Bay
Paul Fleming
Abstract. Layout design and wake steering through wind plant control are each important and complex components in the design and operation of modern wind plants. They are currently optimized separately, but as more and more wind plants implement wake steering as their primary form of operation, there are increasing needs from industry and regulating bodies to combine the layout and control optimization in a co-design process. However, combining these two optimization problems is currently infeasible due to the excessive number of design variables and the very large solution space. In this paper we present a revolutionary method that enables the coupled optimization of wind plant layout and wake steering with no additional computational expense than a traditional layout optimization. This is accomplished through the development of a geometric relationship between turbines to find an approximate optimal yaw angle, bypassing the need for either a nested or coupled wind plant control optimization. The method we present in this paper provides a significant and immediate improvement to wind plant design by enabling the co-design of turbine layout and yaw control for wake steering. A small co-designed plant shown in this paper produces 0.8 % more energy than its sequentially designed counterpart, and we expect larger comparative gains for larger plants with more turbines. This additional energy production comes with no additional infrastructure, turbine hardware, or control software; it is a free consequence of optimizing the turbine layout and yaw control together, resulting in millions of dollars of additional revenue for the wind plants of the future.
Andrew P. J. Stanley et al.
Status: final response (author comments only)
-
CC1: 'Comment on wes-2023-1', Adam Stock, 15 Mar 2023
The paper is a useful and interesting study, definitely worth sharing with the community, but some aspects could perhaps be clarified further within the text, particularly when it comes to the limitations of the work presented.
- It would be interesting to hear the author’s views on the real-world applicability of the methodology. The control codesign methodology appears to only consider the energy capture of the turbines and does not consider other implications of a layout design such as accessibility, water depth or cabling lengths and layouts. Whilst one would not expect these aspects to be fully costed in and included in the analysis, the absence of these factors should be noted in the text with perhaps a comparison in the order of magnitude difference that such aspects may impart. Would these factors be easy to accommodate within the methodology? Is this a topic for future work perhaps?
- The text notes that “we expect larger comparative gains for larger plants with more turbines” but the basis for this assertion is not presented. Is it not the case that, beyond a certain point, wakes can become somewhat of a “wake soup” and there could be diminishing returns? Some discussion of the implications of applying the method on farms with more turbines would be useful.
- The equation for the wake expansion r_wake = 0.1x + r_turbine seems broadly sensible, but there is no reference for this equation. Might this value change in different atmospheric conditions? Does the wake continue expanding indefinitely? Furthermore, it would be useful for the authors to comment on the sensitivity of their results to changes in the wake expansion factor.
- Alongside the result of “0.8% more energy than its sequentially designed counterpart” it would be informative to know the improvement compared to purely optimising layout with no WFFC applied.
- Why was the particular wind rose that was used chosen? Similar to the wake expansion point above, what sort of sensitivity does the method have to different wind roses? As a wider point, the work presents a single example – can the authors be sure that the result is typical? What are the bounds of that “typicality”?
Overall, the work is a very interesting and well set out study that was enjoyable to read. The points above mainly centre around reflecting on the limitations of the work, the impact these limitations have on the conclusions that can be drawn, and the scope for future work that could expand the topic further. The first point above, in particular, I feel is an essential addition to the work to properly place the results within the wider context of optimisation of wind farm layouts.
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-2023-1-CC1 - AC1: 'Reply on CC1', Andrew P.J. Stanley, 18 May 2023
-
RC1: 'Comment on wes-2023-1', Anonymous Referee #1, 02 Apr 2023
- AC2: 'Reply on RC1', Andrew P.J. Stanley, 18 May 2023
-
RC2: 'Comment on wes-2023-1', Sebastian Sanchez Perez-Moreno, 04 Apr 2023
General comments
- This paper provides a novel approach to co-optimise the wind turbine layout and yaw angles for wake steering. This method greatly accelerates the co-design process by avoiding nested optimisation loops that are computationally very expensive. Together with the understanding of the dependency of the optimal yaw angle with respect to the relative position of the nearest waked wind turbine, these are the biggest contributions from this research. The outcome of this paper is directly applicable in engineering processes in our industry.
- It is enourmously appreciated the transparency and repeatability that comes with the code being publicly available. Also want to congratulate the authors on the high quality of the figures, which communicate the outcomes more clearly.
- I would like to read follow up research ideas stemming from this first step, to reduce the uncertainty and risks associated with the assumptions made here (power density, wake model). With more research, industry would deploy this method with more confidence, as the layouts optimised with co-design perform worse than without it, when not operated with wake steering.
- Additionally, could the authors mention something with respect to the turbine fatigue loads expected from implementing these yaw angles?
Specific comments- Line 2: it says "...currently optimized separately... more and more wind plants
implement wake steering as their primary form of operation" (gives impression that wake steering is already implemented in wind plants in operation, is this true?) - Line 54: "...per wind condition". What is meant here, wind speed and direction? Can you be more specific?
- Figure 1: Don't the number of wind speed bins count? Or is a single wind speed per direction used to make this figure? This could be stated explicitly. Over 150 hours for a "small" wind farm is already too long for a single wind speed.
- Line 75: In my opinion, it's somewhat exaggerated the statement that this new approach alone can accelerate the deployment of wind energy and reach goals.
- Line 85: All downstream turbines are waked (if by small amounts) according to Gaussian models, so it would good to be explicit say what model is used here (looks like a top hat Jensen profile). And can you mention what wake decay factor corresponds rougly to the wake radius formula? I am curious about how "optimistic" this wake radius is and how the conclusions can change depending on this wake expansion factor.
- Fig 2c: are there orange or blue coloured points underneath the black dots? Are there clearly many more black dots than coloured? How did you determine the 1D threshold?
- Line 121: why do you re-optimise the yaw angle after a layout has been optimised with co-design? What are the quantitative improvements before this step?
- Do the final re-optimised yaw values correlate nicely with the deterministic yaw angles found during the co-optimisation?
- Line 115: doesn't SLSQP require multiple initial conditions to get closer the global optimum? What were the initial conditions (layout) in 3.1?
- What is the turbine nameplate capacity in example 3.2? What is the plant power density? 2 km x 2 km seems small even for a 3.5 MW WTG rated capacity (14 MW/km2). Is the AEP increase of co-design as big as 0.8% for sites with smaller but more realistic power density (e.g. 5 MW/km2)?
- Does wake steering optimisation for existing plants still have any value after finding this geometric relationship? What's the trade-off between "slow" optimisation and finding the optimal angles deterministically?
Technical corrections- Figure 1: shouldn’t it say "with 24 wind direction bins" instead of "with the 24 wind direction bins"?
- Line 115: typo - should say "optimizer" where it says "optimzer"
Citation: https://doi.org/10.5194/wes-2023-1-RC2 - AC3: 'Reply on RC2', Andrew P.J. Stanley, 18 May 2023
Andrew P. J. Stanley et al.
Andrew P. J. Stanley et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 314 | 163 | 19 | 496 | 3 | 1 |
- HTML: 314
- PDF: 163
- XML: 19
- Total: 496
- BibTeX: 3
- EndNote: 1
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1