the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Aeroelastic Tailoring of Wind Turbine Rotors Using High-Fidelity Multidisciplinary Design Optimization
Abstract. Reducing the cost of energy from wind power is a critical step towards decarbonizing the electric grid and mitigating climate change effects. Computational models are crucial in understanding the complex, multiphysics interactions of modern, highly flexible wind turbine rotors. When coupled with numerical optimization, such models provide an efficient way to explore the design space. We propose a high-fidelity aerostructural framework that couples computational fluid dynamics with 5 computational structural mechanics to analyze and optimize wind turbines. The framework uses a gradient-based optimization strategy with gradients efficiently computed using a coupled-adjoint approach. We optimize a benchmark utility-scale wind turbine rotor and explore its trade-offs between steady-state aerodynamic efficiency and structural weight. The optimizations account for a representative below-rated operating condition and use more than 100 structural and geometric design variables. The monolithic approach we propose is compared with a loosely-coupled optimization strategy used for the structural sizing 10 of the baseline rotor layout. We then discuss specific optimized blade features and a broader design trade space exploration between extracted torque and rotor mass. Optimized layouts increase the torque by up to 14 % and reduce the mass by up to 9 % or reduce the mass by up to 27 % with the same torque output compared to the baseline blade. The results demonstrate the benefits of optimizing a tightly-coupled aerostructural model and reveal additional insights that high-fidelity analysis provide to complement conventional design approaches.
- Preprint
(2229 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
CC1: 'Comment on wes-2023-10', Georg Raimund Pirrung, 28 Feb 2023
Dear authors,
thank you very much for writing this interesting article.
I have a minor comment on the reference to Horcas et al. (2022).
You write 'Horcas et al. (2022) showed how a tool based on the same physical models of OpenFAST, HAWC2, overpredicts
the loads on a 10 MW turbine at below-rated wind speeds and underestimates the benefits of curved wing tips, even for a
steady power curve.'It is true that HAWC2, similar to OpenFAST, has a BEM model implemented. The model, including a correction of the BEM equations for curved blade shapes, is described here: https://wes.copernicus.org/articles/5/1/2020/ .
However, many comparisons have shown that different BEM implementations can generate results that differ significantly between implementations. One of the latest comparisons that includes both OpenFAST and HAWC2 can be found here: https://wes.copernicus.org/articles/8/211/2023/
As shown in that comparison, the differences between different BEM implementations can be quite large even for turbines with straight blades.
And the differences will increase when turbines with curved blades or curved tips are studied. This was also shown in Horcas et al. (2022) where BEM models with and without radial induction models were compared.In addition to the low-fidelity BEM model and the high-fidelity CFD model, also two mid-fidelity models were used in Horcas et al. (2022): A combination of a near wake and vortex cylinder model that was implemented in HAWC2 and a Lifting Line model, which both lead to improved results compared to the BEM model.
Maybe you could extend the sentence about Horcas et al. (2022) slightly to also reflect that different BEM model implementations can generate different results, and that models covering a range of aerodynamic fidelities have been evaluated in Horcas et al (2022)?
Thank you very much for considering this comment.
Best regards,
Georg PirrungDisclaimer: 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-2023-10-CC1 - AC1: 'Reply on CC1', Marco Mangano, 17 Apr 2023
-
RC1: 'Comment on wes-2023-10', Anonymous Referee #1, 09 Mar 2023
The study presents a tightly coupled, 3D optimization framework based on FEM analysis and CFD. The method has a lot of potential and if properly fine-tuned can be used to greatly improve the current state o the art when it come to wind turbine rotor design.
The work is very interesting, and this line of work has a lot of potential. Nevertheless, some of the details of this work raise some questions in my opinion. Firstly, the scope of this work is not clear: is it intended as an illustration and first showcasing of the coupled optimization method? If so important details regarding the method’s details are missing throughout the paper, and it would be very hard for a third party to reproduce the results showcased with the information contained in the paper. On the other hand, if the objective is to discuss the results of the optimization discussed in the paper, various corners appear to have been cut: the blade is made of aluminum and no details regarding how the distribution along the blade of the various thickness panels is chosen is given. From an aerodynamic perspective, no mesh convergence & important details on model set-up are presented, and baseline results are not compared to other author’s predictions for this testcase. Moreover, the single-point optimization, without accounting for extreme loads in other operating and parked conditions is questionable, and the authors also acknowledge this in the paper.
- AC2: 'Reply on RC1', Marco Mangano, 18 Apr 2023
-
RC2: 'Comment on wes-2023-10', Pietro Bortolotti, 18 Mar 2023
The paper describes cutting-edge research in the field of high-fidelity design optimization of wind turbine rotors. The numerical framework described in the paper is unique and worth a publication. The authors have done an impressive amount of work and the article reads very well (although I prefer shorter and more concise papers).
This said, I am left a little puzzled by what a reader can learn here.
The numerical framework is described in several other publications and this new article seems more focused on results. However I struggle to extract takeaways that can be applied to the real world.
First, blades are modeled as made of Aluminum, which is clearly not the case for real blades. This to me means that little to nothing of the structural results can be applied to real blades made of composites.
If we stick to aerodynamics, I fear that the proposed design solutions violate standard stall margins, especially along span.
Overall, I really miss a comparison against lower fidelity conventional tools. Was the whole high-fidelity framework really necessary to identify these tradeoffs? Why wouldn’t the same tradeoffs emerge at lower fidelity? As a reader, I’m left with the strong feeling that the large mass reductions are too good to be true and only a numerical artifact generated by the unresolved weaknesses of this novel framework. Can the authors prove me wrong?
A list of additional minor comments is provided below:
Page 1 Abstract: It would be nice to learn how the optimizer achieves such large gains already in the abstract
Page 3 Line 53-60: The whole first paragraph can be safely deleted. No need to talk about climate change in a wind energy journal
Page 3 Line 65: lower solidity does not necessarily mean higher efficiency. For modern WTs, lower solidity mostly means higher TSR and lower drivetrain costs
Page 4 Line 108: OpenFAST is not really based on a multibody dynamics model
Page 4 Line 113: A reference seems missing. Maybe some publications from IEA Wind TCP Tasks 29 or 47 could be used
Page 5 Line 120: I would add “unless coupled to an airfoil solver”
Page 10 Line 252: the fact that blades are made of Aluminum is a major limitation of this study. This shortcoming cannot be buried in page 10 or many readers will miss it.
Page 13 Line 344: Why 14% more thrust? +14% in max thrust would be a big increase. At the same time +14% in region II might not be an issue as long as the peak thrust shaving logic is applied when thrust reaches its maximum. Overall more discussion around this assumption seems needed.
Page 14 Line 355: In the report DTU Wind Energy Report-I-0092 rated TSR is 7.5, not 7.8. Who is wrong?
Page 20 Figure 6: Nowhere in the paper I see the word stall, which is however a key consideration in real wind turbine blade design. Designers must enforce a minimum margin to stall, or the turbulent inflow will keep the blade in and out of stall. If you had this constraint, I doubt you’d be able to drop twist so much in the mid span. Please check the operating angles of attack of the airfoils and how close you are to the stall point
Page 21 Figure 7: Max chord is often constrained. Here different solutions have different max chord values. Is this a fair comparison? Why doesn’t the optimizer favor larger chords?
Citation: https://doi.org/10.5194/wes-2023-10-RC2 - AC3: 'Reply on RC2', Marco Mangano, 18 Apr 2023
Status: closed
-
CC1: 'Comment on wes-2023-10', Georg Raimund Pirrung, 28 Feb 2023
Dear authors,
thank you very much for writing this interesting article.
I have a minor comment on the reference to Horcas et al. (2022).
You write 'Horcas et al. (2022) showed how a tool based on the same physical models of OpenFAST, HAWC2, overpredicts
the loads on a 10 MW turbine at below-rated wind speeds and underestimates the benefits of curved wing tips, even for a
steady power curve.'It is true that HAWC2, similar to OpenFAST, has a BEM model implemented. The model, including a correction of the BEM equations for curved blade shapes, is described here: https://wes.copernicus.org/articles/5/1/2020/ .
However, many comparisons have shown that different BEM implementations can generate results that differ significantly between implementations. One of the latest comparisons that includes both OpenFAST and HAWC2 can be found here: https://wes.copernicus.org/articles/8/211/2023/
As shown in that comparison, the differences between different BEM implementations can be quite large even for turbines with straight blades.
And the differences will increase when turbines with curved blades or curved tips are studied. This was also shown in Horcas et al. (2022) where BEM models with and without radial induction models were compared.In addition to the low-fidelity BEM model and the high-fidelity CFD model, also two mid-fidelity models were used in Horcas et al. (2022): A combination of a near wake and vortex cylinder model that was implemented in HAWC2 and a Lifting Line model, which both lead to improved results compared to the BEM model.
Maybe you could extend the sentence about Horcas et al. (2022) slightly to also reflect that different BEM model implementations can generate different results, and that models covering a range of aerodynamic fidelities have been evaluated in Horcas et al (2022)?
Thank you very much for considering this comment.
Best regards,
Georg PirrungDisclaimer: 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-2023-10-CC1 - AC1: 'Reply on CC1', Marco Mangano, 17 Apr 2023
-
RC1: 'Comment on wes-2023-10', Anonymous Referee #1, 09 Mar 2023
The study presents a tightly coupled, 3D optimization framework based on FEM analysis and CFD. The method has a lot of potential and if properly fine-tuned can be used to greatly improve the current state o the art when it come to wind turbine rotor design.
The work is very interesting, and this line of work has a lot of potential. Nevertheless, some of the details of this work raise some questions in my opinion. Firstly, the scope of this work is not clear: is it intended as an illustration and first showcasing of the coupled optimization method? If so important details regarding the method’s details are missing throughout the paper, and it would be very hard for a third party to reproduce the results showcased with the information contained in the paper. On the other hand, if the objective is to discuss the results of the optimization discussed in the paper, various corners appear to have been cut: the blade is made of aluminum and no details regarding how the distribution along the blade of the various thickness panels is chosen is given. From an aerodynamic perspective, no mesh convergence & important details on model set-up are presented, and baseline results are not compared to other author’s predictions for this testcase. Moreover, the single-point optimization, without accounting for extreme loads in other operating and parked conditions is questionable, and the authors also acknowledge this in the paper.
- AC2: 'Reply on RC1', Marco Mangano, 18 Apr 2023
-
RC2: 'Comment on wes-2023-10', Pietro Bortolotti, 18 Mar 2023
The paper describes cutting-edge research in the field of high-fidelity design optimization of wind turbine rotors. The numerical framework described in the paper is unique and worth a publication. The authors have done an impressive amount of work and the article reads very well (although I prefer shorter and more concise papers).
This said, I am left a little puzzled by what a reader can learn here.
The numerical framework is described in several other publications and this new article seems more focused on results. However I struggle to extract takeaways that can be applied to the real world.
First, blades are modeled as made of Aluminum, which is clearly not the case for real blades. This to me means that little to nothing of the structural results can be applied to real blades made of composites.
If we stick to aerodynamics, I fear that the proposed design solutions violate standard stall margins, especially along span.
Overall, I really miss a comparison against lower fidelity conventional tools. Was the whole high-fidelity framework really necessary to identify these tradeoffs? Why wouldn’t the same tradeoffs emerge at lower fidelity? As a reader, I’m left with the strong feeling that the large mass reductions are too good to be true and only a numerical artifact generated by the unresolved weaknesses of this novel framework. Can the authors prove me wrong?
A list of additional minor comments is provided below:
Page 1 Abstract: It would be nice to learn how the optimizer achieves such large gains already in the abstract
Page 3 Line 53-60: The whole first paragraph can be safely deleted. No need to talk about climate change in a wind energy journal
Page 3 Line 65: lower solidity does not necessarily mean higher efficiency. For modern WTs, lower solidity mostly means higher TSR and lower drivetrain costs
Page 4 Line 108: OpenFAST is not really based on a multibody dynamics model
Page 4 Line 113: A reference seems missing. Maybe some publications from IEA Wind TCP Tasks 29 or 47 could be used
Page 5 Line 120: I would add “unless coupled to an airfoil solver”
Page 10 Line 252: the fact that blades are made of Aluminum is a major limitation of this study. This shortcoming cannot be buried in page 10 or many readers will miss it.
Page 13 Line 344: Why 14% more thrust? +14% in max thrust would be a big increase. At the same time +14% in region II might not be an issue as long as the peak thrust shaving logic is applied when thrust reaches its maximum. Overall more discussion around this assumption seems needed.
Page 14 Line 355: In the report DTU Wind Energy Report-I-0092 rated TSR is 7.5, not 7.8. Who is wrong?
Page 20 Figure 6: Nowhere in the paper I see the word stall, which is however a key consideration in real wind turbine blade design. Designers must enforce a minimum margin to stall, or the turbulent inflow will keep the blade in and out of stall. If you had this constraint, I doubt you’d be able to drop twist so much in the mid span. Please check the operating angles of attack of the airfoils and how close you are to the stall point
Page 21 Figure 7: Max chord is often constrained. Here different solutions have different max chord values. Is this a fair comparison? Why doesn’t the optimizer favor larger chords?
Citation: https://doi.org/10.5194/wes-2023-10-RC2 - AC3: 'Reply on RC2', Marco Mangano, 18 Apr 2023
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
1,005 | 317 | 45 | 1,367 | 35 | 30 |
- HTML: 1,005
- PDF: 317
- XML: 45
- Total: 1,367
- BibTeX: 35
- EndNote: 30
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1