the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Feb 2022
Quantifying the effect of blockage for wind farm layout optimization
Abstract. Wind plant blockage is a phenomenon where the presence of a large wind farm creates a disturbance to air flow external to the farm itself that is not accounted for by wake effects. This is typically manifested as a slowdown in wind velocity upstream from a plant, which can be shown to decrease power production. Knowing a priori how a wind farm's layout will generate blockage could help to improve the accuracy of AEP calculations and suggest more blockage-optimal turbine layouts. In this study, we consider the effect of wind plant blockage and perform multiple types of layout optimization to reduce blockage while solving the flow physics using computational fluid dynamics. We present a variety of methods to quantify this blockage effect, ranging from localized measurements taken in the proximity of each turbine to farm-wide integral measurements designed to capture the velocity decrease in a more global way. We then investigate each blockage metric using simple case studies designed to isolate effects due to layout changes. While all metrics show sensitivity during this testing, the integral metrics better avoid spurious effects due to waking and are better suited to evaluating blockage which we find to be a cumulative effect. We then perform multiple farm layout optimizations using these blockage metrics as objectives and find that in the absence of a power production constraint the optimized layouts tend to minimize disturbance to the surrounding flow by creating streamwise rows of turbines. Finally, we present a power-constrained layout optimization which explicitly illustrates the trade-off between designing for blockage and power. This work presents a variety of different blockage definitions and offers a set of recommendations regarding their deployment, expected behavior, and viability for consideration as part of layout planning.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(1740 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on wes-2022-7', Luca Lanzilao, 28 Feb 2022
The aims of this paper are twofold. First, the authors try to identify the most efficient and physically relevant way to estimate flow blockage in the front of a farm. To this end, three different metrics are adopted: a single point, resembling a met mast, a box and finally a cylinder. These regions extend both upstream and above the first row of turbines. The flow blockage is then measured while comparing the velocity fields with and without turbines. Experiments extend to a large variety of farm layouts. The authors conclude that in the majority of the cases and with all metrics adopted, the flow slows down in front of the farm and accelerates above it. Next, an optimization problem is formulated which aims to minimize the velocity slows down upstream of the farm and the velocity speed-up above the first row of turbines. Optimal wind-farm layouts including blockage effects only and blockage plus a minimal power constraint are then suggested. I believe this paper is of interest to the wind energy community. Moreover, it contains good quality figures. However, I also think that some important information is currently missing, such as the numerical domain description, gradient and model validation and a grid sensitivity study. These additions could increase the strength of the results presented. Here below, you can find some scientific questions and technical comments.
Scientific comments/questions
- Introduction: some more context would be appreciated. What do you mean when you refer to blockage? Is this related to a cumulative turbine induction effect (also called hydrodynamic effect), or is it also related to atmospheric conditions (such as flow blockage induced by atmospheric stability, for instance)? I would consider adding some more key points on how this work differs from previous ones. Is this the first of its kind? The work from Lanzilao and Meyers (https://doi.org/10.5194/wes-6-247-2021) also deals with optimization including blockage effects. How the current work differs from that one (for instance)?
- Section 2.1: How is the spatially-varying eddy viscosity computed? Which are the boundary conditions imposed at the sides and top of the domain? How is the presence of the wall treated (bottom boundary condition)? Overall, I would like to see some more information on the fluid solver.
- Section 2.2: Several studies that adopt the adjoint method to efficiently compute the gradient also show its accuracy using a comparison with a finite difference approximation (see Munters and Meyers (https://doi.org/10.3390/en11010177), Yilmaz and Meyers (https://doi.org/10.1063/1.5038600) or Lanzilao and Meyers (https://doi.org/10.5194/wes-6-247-2021), for instance). You mention in line 63 that the gradient you obtain is accurate. How accurate is it? Have you done such verification? This could build some additional thrust in the optimal layouts shown in section 4.
- Line 60: I would appreciate a more clear definition of the state and control variables adopted in your optimization framework.
- Line 61: Do you use dolphin-adjoint or SNOPT? It is not clear the relationship between these two packages to me, but it could be since I’m not familiar with these.
- Section 3: Table 1 reports detailed information about the turbine specifications and inflow conditions. However, I have not found information about the computational domain and grid resolution. On which basis were these chosen? I am asking this because the use of a small domain could eventually distort the farm induced blockage due to the close presence of boundaries, on which boundary conditions are imposed. Also, have you performed a grid sensitivity study? It would be interesting to see how J depends on the grid resolution (both horizontal and vertical).
- Line 151: Have you checked whether this behavior is also observed when placing the observation point differently?
- Line 155: It is not clear to me why the measurements within the cylinder would also include wake effects. Could you comment on this?
- Figure 2: Would it be possible to express the blockage not in m/s but rather as a percentage of the inflow velocity? This will give a better idea of the magnitude of such an effect. The same applies to all other figures. Also, how do the measured blockage values compare to other studies in the literature (such as Bleeg et al (doi:10.3390/en11061609), or Segalini and Dahlberg (doi:10.1002/we.2413))? If possible, it would be nice to make some connections.
- Figure 3: for the upstream point, the blockage becomes positive. This means that the presence of the farm causes a flow speed up in the upstream region. Could you comment on this?
- General comments about section 3:
- In all three sensitivity studies on the wind-farm layout carried out in this section, the upstream cylinder always predicts a blockage that is two or three orders of magnitude higher than the one measured with the upstream box or upstream point metrics. In some cases, this metric predicts a flow blockage close to 50%, which is rather far from the values usually seen in the literature (when referring to hydrodynamic blockage only). Could you further comment on this?
- In this study, you have fixed the 3 metrics and varied the wind-farm layout extensively. What about the vice versa? That is, fixing the farm layout and varying the position and shape of the three metrics. Could this provide additional insights on how to best estimate flow blockage?
- Out of curiosity, have you tried to do a similar sensitivity study when changing the rotor diameter? It would be nice to see how blockage scales with it. However, I understand that this might be out of the scope of the current study.
- It would be nice to see at least one 2D visualization of the velocity and/or pressure fields (top and/or side view). This would provide a better understanding of how turbulent structures and turbine wakes are resolved with the fluid solver adopted.
- Section 4.1: Wake effects induce much higher power losses than blockage effects. Therefore, what would be a real-case application of such a case study? I think it is very nice to know which farm layout minimizes blockage, but I feel that at the same time this farm layout must also account for turbine wakes.
- Line 220: I understand what you mean with this description of the optimization problem. However, I think the article would benefit from a more rigorous mathematical description of it. For instance, writing “minimize/maximize” could be ambiguous to a reader who didn’t go through the full manuscript.
- Figure 5: The difference in blockage between the initial and optimized wind-farm layout is minimal in the upstream box case. On the other hand, higher speed recoveries are obtained when using the upstream point or upstream cylinder metrics. Could you comment on this?
- Line 305: are you also planning to include buoyancy forces in your governing equations? In such a case, the flow will be re-directed both above and at the side of the farm, depending on the strength and height of an eventual capping inversion. Moreover, atmospheric stability has an impact on the reduction in velocity upstream of the farm (see Allaerts and Meyers (https://doi.org/10.1007/s10546-017-0307-5) and Schneemann et al (https://doi.org/10.5194/wes-6-521-2021)). Would it be possible to take these effects into account with the proposed optimization technique?
Technical comments
- Line 83: Typo in referencing the figure
- Table 1: Typo in second line “HH” -> H
- Line 163: typo “the the” -> the (end of the line)
- Line 173: typo “trend” -> tend
Citation: https://doi.org/10.5194/wes-2022-7-RC1 -
RC2: 'Comment on wes-2022-7', Anonymous Referee #2, 24 Mar 2022
The authors are trying to address the issue of wind farm blockage by defining some new blockage metrics and using a finite-difference based adjoint-solver in conjunction with a CFD model to optimize a wind farm layout with respect to blockage. Whilst the topic and methods employed by the authors are extremely relevant to wind farm optimization, the overall structure, quality and scientific value of the paper does not justify its publication and should be rejected. This is not to say that an improved version of the paper should not be reconsidered, however the necessary changes would probably lead to a very different paper. A resubmission should be considered after addressing some major flaws of the current version addressed in the following paragraphs.
To me it is currently unclear what the authors are trying to answer. Please try to formulate a research hypothesis and structure your paper around it. Currently, it is unclear what question the authors are trying to answer, is it whether a metric can be devised for blockage or whether one can optimize the wind farm layout for blockage or is it the optimization framework which is being verified? If they are trying to optimize the wind farm why would one use the metrics proposed by the authors and not directly optimize for power? Why for a single wind direction and wind speed? Afterall blockage is a very small effect – as the authors are also realizing once they start to optimize for power directly – so why should one optimize for a measure of wind speed upstream of the wind farm? This would require an upstream wind speed measurement to give a direct relationship to the power production in a wind farm, but this is not the case as pointed out by many other researchers working on blockage (maybe not spotted by the authors as the reference list is rather limited). The larger blockage at the most upstream turbine also means that less power is extracted by the first row of turbines, energy that can then be extracted by downstream turbines in the wind farm, very similar as in induction control of farms.
A major problem with the optimization is not the procedure itself, that is rather interesting and holds large potential - and maybe that is where the focus should have been - but the metrics devised by the authors. They do not sufficiently justify and verify their choice, why should one use these measures. Why not the disc-averaged velocity, a much more relevant quantity with respect to power production, the ultimate goal of any wind farm optimization (ignoring constraints like water depth, cable length …)? The flow upstream of a turbine is always just a proxy for power production, however it is the induction in the rotor plane that matters. The velocity upstream is not a good indicator in wind farms or any location where we have flow evolution.
Finally, the authors need to work on the reproducibility of their work. They do not even state explicitly that they are solving the RANS equations. There is no numerical domain description or methodology, no grid study, no explanation of the Gaussian kernels they use to represent the ADs. It is also unclear how they are setting the thrust at the rotor inside the wind farm? If the thrust coefficient of downstream turbines is actually influenced by upstream turbines (the case when not using constant CT’ everywhere) then it is hard to conclude on what the effect is of adding downstream turbines as this will completely depend on the turbine type and wind speed as CT is changing between turbines. If it is not constant it is hard to derive any general take-aways from the reported metric evolution with number of turbines. It also remained unclear why the authors chose to optimize for a single wind direction and wind speed (the description is very limited so actually unsure if this was the case)?
Citation: https://doi.org/10.5194/wes-2022-7-RC2 -
AC1: 'Comment on wes-2022-7', Ethan Young, 16 Jun 2022
After discussing the reviewer responses with my co-authors, we have decided to retract the current manuscript. The reviews highlighted the need for major revisions and proposed interesting new work and, unfortunately, we do not currently have the resources to implement these changes in a timely manner nor can we reasonably expect to complete the work with a short-term extension of the deadline. We fully plan on revising the paper and resubmitting it once we have improved its quality.
Citation: https://doi.org/10.5194/wes-2022-7-AC1
Interactive discussion
Status: closed
-
RC1: 'Comment on wes-2022-7', Luca Lanzilao, 28 Feb 2022
The aims of this paper are twofold. First, the authors try to identify the most efficient and physically relevant way to estimate flow blockage in the front of a farm. To this end, three different metrics are adopted: a single point, resembling a met mast, a box and finally a cylinder. These regions extend both upstream and above the first row of turbines. The flow blockage is then measured while comparing the velocity fields with and without turbines. Experiments extend to a large variety of farm layouts. The authors conclude that in the majority of the cases and with all metrics adopted, the flow slows down in front of the farm and accelerates above it. Next, an optimization problem is formulated which aims to minimize the velocity slows down upstream of the farm and the velocity speed-up above the first row of turbines. Optimal wind-farm layouts including blockage effects only and blockage plus a minimal power constraint are then suggested. I believe this paper is of interest to the wind energy community. Moreover, it contains good quality figures. However, I also think that some important information is currently missing, such as the numerical domain description, gradient and model validation and a grid sensitivity study. These additions could increase the strength of the results presented. Here below, you can find some scientific questions and technical comments.
Scientific comments/questions
- Introduction: some more context would be appreciated. What do you mean when you refer to blockage? Is this related to a cumulative turbine induction effect (also called hydrodynamic effect), or is it also related to atmospheric conditions (such as flow blockage induced by atmospheric stability, for instance)? I would consider adding some more key points on how this work differs from previous ones. Is this the first of its kind? The work from Lanzilao and Meyers (https://doi.org/10.5194/wes-6-247-2021) also deals with optimization including blockage effects. How the current work differs from that one (for instance)?
- Section 2.1: How is the spatially-varying eddy viscosity computed? Which are the boundary conditions imposed at the sides and top of the domain? How is the presence of the wall treated (bottom boundary condition)? Overall, I would like to see some more information on the fluid solver.
- Section 2.2: Several studies that adopt the adjoint method to efficiently compute the gradient also show its accuracy using a comparison with a finite difference approximation (see Munters and Meyers (https://doi.org/10.3390/en11010177), Yilmaz and Meyers (https://doi.org/10.1063/1.5038600) or Lanzilao and Meyers (https://doi.org/10.5194/wes-6-247-2021), for instance). You mention in line 63 that the gradient you obtain is accurate. How accurate is it? Have you done such verification? This could build some additional thrust in the optimal layouts shown in section 4.
- Line 60: I would appreciate a more clear definition of the state and control variables adopted in your optimization framework.
- Line 61: Do you use dolphin-adjoint or SNOPT? It is not clear the relationship between these two packages to me, but it could be since I’m not familiar with these.
- Section 3: Table 1 reports detailed information about the turbine specifications and inflow conditions. However, I have not found information about the computational domain and grid resolution. On which basis were these chosen? I am asking this because the use of a small domain could eventually distort the farm induced blockage due to the close presence of boundaries, on which boundary conditions are imposed. Also, have you performed a grid sensitivity study? It would be interesting to see how J depends on the grid resolution (both horizontal and vertical).
- Line 151: Have you checked whether this behavior is also observed when placing the observation point differently?
- Line 155: It is not clear to me why the measurements within the cylinder would also include wake effects. Could you comment on this?
- Figure 2: Would it be possible to express the blockage not in m/s but rather as a percentage of the inflow velocity? This will give a better idea of the magnitude of such an effect. The same applies to all other figures. Also, how do the measured blockage values compare to other studies in the literature (such as Bleeg et al (doi:10.3390/en11061609), or Segalini and Dahlberg (doi:10.1002/we.2413))? If possible, it would be nice to make some connections.
- Figure 3: for the upstream point, the blockage becomes positive. This means that the presence of the farm causes a flow speed up in the upstream region. Could you comment on this?
- General comments about section 3:
- In all three sensitivity studies on the wind-farm layout carried out in this section, the upstream cylinder always predicts a blockage that is two or three orders of magnitude higher than the one measured with the upstream box or upstream point metrics. In some cases, this metric predicts a flow blockage close to 50%, which is rather far from the values usually seen in the literature (when referring to hydrodynamic blockage only). Could you further comment on this?
- In this study, you have fixed the 3 metrics and varied the wind-farm layout extensively. What about the vice versa? That is, fixing the farm layout and varying the position and shape of the three metrics. Could this provide additional insights on how to best estimate flow blockage?
- Out of curiosity, have you tried to do a similar sensitivity study when changing the rotor diameter? It would be nice to see how blockage scales with it. However, I understand that this might be out of the scope of the current study.
- It would be nice to see at least one 2D visualization of the velocity and/or pressure fields (top and/or side view). This would provide a better understanding of how turbulent structures and turbine wakes are resolved with the fluid solver adopted.
- Section 4.1: Wake effects induce much higher power losses than blockage effects. Therefore, what would be a real-case application of such a case study? I think it is very nice to know which farm layout minimizes blockage, but I feel that at the same time this farm layout must also account for turbine wakes.
- Line 220: I understand what you mean with this description of the optimization problem. However, I think the article would benefit from a more rigorous mathematical description of it. For instance, writing “minimize/maximize” could be ambiguous to a reader who didn’t go through the full manuscript.
- Figure 5: The difference in blockage between the initial and optimized wind-farm layout is minimal in the upstream box case. On the other hand, higher speed recoveries are obtained when using the upstream point or upstream cylinder metrics. Could you comment on this?
- Line 305: are you also planning to include buoyancy forces in your governing equations? In such a case, the flow will be re-directed both above and at the side of the farm, depending on the strength and height of an eventual capping inversion. Moreover, atmospheric stability has an impact on the reduction in velocity upstream of the farm (see Allaerts and Meyers (https://doi.org/10.1007/s10546-017-0307-5) and Schneemann et al (https://doi.org/10.5194/wes-6-521-2021)). Would it be possible to take these effects into account with the proposed optimization technique?
Technical comments
- Line 83: Typo in referencing the figure
- Table 1: Typo in second line “HH” -> H
- Line 163: typo “the the” -> the (end of the line)
- Line 173: typo “trend” -> tend
Citation: https://doi.org/10.5194/wes-2022-7-RC1 -
RC2: 'Comment on wes-2022-7', Anonymous Referee #2, 24 Mar 2022
The authors are trying to address the issue of wind farm blockage by defining some new blockage metrics and using a finite-difference based adjoint-solver in conjunction with a CFD model to optimize a wind farm layout with respect to blockage. Whilst the topic and methods employed by the authors are extremely relevant to wind farm optimization, the overall structure, quality and scientific value of the paper does not justify its publication and should be rejected. This is not to say that an improved version of the paper should not be reconsidered, however the necessary changes would probably lead to a very different paper. A resubmission should be considered after addressing some major flaws of the current version addressed in the following paragraphs.
To me it is currently unclear what the authors are trying to answer. Please try to formulate a research hypothesis and structure your paper around it. Currently, it is unclear what question the authors are trying to answer, is it whether a metric can be devised for blockage or whether one can optimize the wind farm layout for blockage or is it the optimization framework which is being verified? If they are trying to optimize the wind farm why would one use the metrics proposed by the authors and not directly optimize for power? Why for a single wind direction and wind speed? Afterall blockage is a very small effect – as the authors are also realizing once they start to optimize for power directly – so why should one optimize for a measure of wind speed upstream of the wind farm? This would require an upstream wind speed measurement to give a direct relationship to the power production in a wind farm, but this is not the case as pointed out by many other researchers working on blockage (maybe not spotted by the authors as the reference list is rather limited). The larger blockage at the most upstream turbine also means that less power is extracted by the first row of turbines, energy that can then be extracted by downstream turbines in the wind farm, very similar as in induction control of farms.
A major problem with the optimization is not the procedure itself, that is rather interesting and holds large potential - and maybe that is where the focus should have been - but the metrics devised by the authors. They do not sufficiently justify and verify their choice, why should one use these measures. Why not the disc-averaged velocity, a much more relevant quantity with respect to power production, the ultimate goal of any wind farm optimization (ignoring constraints like water depth, cable length …)? The flow upstream of a turbine is always just a proxy for power production, however it is the induction in the rotor plane that matters. The velocity upstream is not a good indicator in wind farms or any location where we have flow evolution.
Finally, the authors need to work on the reproducibility of their work. They do not even state explicitly that they are solving the RANS equations. There is no numerical domain description or methodology, no grid study, no explanation of the Gaussian kernels they use to represent the ADs. It is also unclear how they are setting the thrust at the rotor inside the wind farm? If the thrust coefficient of downstream turbines is actually influenced by upstream turbines (the case when not using constant CT’ everywhere) then it is hard to conclude on what the effect is of adding downstream turbines as this will completely depend on the turbine type and wind speed as CT is changing between turbines. If it is not constant it is hard to derive any general take-aways from the reported metric evolution with number of turbines. It also remained unclear why the authors chose to optimize for a single wind direction and wind speed (the description is very limited so actually unsure if this was the case)?
Citation: https://doi.org/10.5194/wes-2022-7-RC2 -
AC1: 'Comment on wes-2022-7', Ethan Young, 16 Jun 2022
After discussing the reviewer responses with my co-authors, we have decided to retract the current manuscript. The reviews highlighted the need for major revisions and proposed interesting new work and, unfortunately, we do not currently have the resources to implement these changes in a timely manner nor can we reasonably expect to complete the work with a short-term extension of the deadline. We fully plan on revising the paper and resubmitting it once we have improved its quality.
Citation: https://doi.org/10.5194/wes-2022-7-AC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 579 | 250 | 33 | 862 | 36 | 30 |
- HTML: 579
- PDF: 250
- XML: 33
- Total: 862
- BibTeX: 36
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1