the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Nov 2025
Integrated Control of Floating Offshore Wind Farms with Reconfigurable Layouts
Abstract. This study proposes an integrated control method for floating offshore wind farms (FOWFs) that seeks to maximize farm-level power output or regulate it to a prescribed reference while mitigating wake-induced losses. To achieve these objectives, the method integrates existing control strategies: turbine repositioning, wake steering, power derating, and dynamic wake mixing, within a unified framework that adaptively selects the most effective combination based on wind conditions and control goals. This integration is motivated by the fact that individual strategies may be effective only under specific conditions or broadly effective but not always optimal, whereas their coordinated use can deliver robust performance improvements across a broad range of operating scenarios. The framework targets FOWFs with reconfigurable layouts, where turbines are mounted on floating platforms anchored to the seabed with sufficiently long and slack mooring lines, allowing them to shift within a certain range and thereby enabling controlled positional adjustments. Numerical simulations using the flow redirection and induction in steady state (FLORIS) engineering wake model show that the integrated method consistently outperforms any individual strategy. These findings highlight the potential of integrated control to enhance the efficiency, flexibility, and adaptability of FOWFs, offering a promising pathway to overcome the limitations and improve the performance of standalone control methods.
- Preprint
(8415 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on wes-2025-240', Anonymous Referee #1, 12 Dec 2025
-
RC2: 'Comment on wes-2025-240', Anonymous Referee #2, 18 Dec 2025
Integrated Control of Floating Offshore Wind Farms with Reconfigurable Layouts
VERDICT
Overall, I like the paper and the contribution it makes to the field. I do, however, feel that the methodology needs to be expanded to specify in more detail all the elements that are included in the proposed framework. In particular, it remains unclear which dynamic wake mixing method is used. In addition, as also noted by the other reviewer, the paper heavily relies on FLORIS. While FLORIS is fundamentally a steady-state engineering model, it could indeed be possible that some dynamic effects (e.g., dynamic wake mixing) are approximated, simulated, or evaluated within your setup. If so, this requires a detailed justification in the methodology section.I also miss a discussion on the presumed constant yaw angle that is required when relocating a turbine from its initial position, and the possible steady and periodic structural loads this introduces. Furthermore, you assume a hub-height wind speed measurement; would a rotor-effective wind speed (REWS) not be a better representation, especially given increasing turbine sizes?
I also do not understand the choice for the multi-objective NSGA-II optimizer; you appear to have a single objective.
I like the high-quality, self-created figures, but please include more elaborate captions underneath them, explaining what can be seen and what can be concluded.
Overall, this is a paper with clear potential, but the methodology in particular needs to be expanded, with stronger justification of the modeling and optimization choices that were made. I therefore suggest a major revision.
COMMENTS
-- General
- Throughout the paper: include more elaborate figure captions. Clearly state what is shown, what the remarkable elements are, and what conclusions can be drawn.-- Abstract
- What does “regulate it to a prescribed reference” mean? Does this refer to total farm power?
- I also find the description of the integrated/unified framework unclear. Only at the end of the introduction does it become apparent that this is an optimization framework. After reading the first sentences of the abstract, it is not clear what the work entails. These sentences form the core of the abstract but currently lack clarity.
- What is the input to your framework (reference/setpoint/demand), and what are its outputs? Please make this explicit.
- If a turbine is relocated from its initial position via yaw misalignment, this yaw misalignment must be maintained to compensate for the restoring force of the mooring lines. I believe the introduction should already mention the additional rotor and structural loads (both steady and periodic) that this induces.
- It remains unclear in both the abstract and introduction what type of “dynamic wake mixing” is intended. Is this dynamic induction using a sinusoidal, helical, or other approach? This should be clarified early in the paper.-- Section 2
- You assume access to the free-stream wind speed at hub height. Would it not be more appropriate to use a rotor-effective wind speed estimate (REWS) rather than a single hub-height measurement? You could potentially alleviate this assumption by estimating wind speed using a REWS estimator. Please comment on this and consider referencing relevant literature.
- Consider renaming “power regulation” to “power derating” throughout the paper.-- Section 3
- “βamp,i (deg), the wake mixing intensity of turbine i”: what exactly does this represent? Is this related to dynamic induction control, where βamp is the amplitude of a sinusoidal yaw signal, or to a helical implementation? Please clarify.
- “The wind comes from the right, representing the worst case with a relative angle of 60 degrees.” Left or right? Please double-check, and also clarify the 60-degree relative angle.
- Again, how are the structural loads associated with such large turbine–wind misalignments accounted for or justified?
- Fig. 6: I do not understand the last subfigure, where you state that there is a 60-degree relative angle. The turbine appears rotated by 180 degrees, and the wind now comes from the right. What is the difference compared to the second subfigure?
- Eq. (5): You could omit one of the weights (e.g., normalize one to unity) and only optimize the other, which would simplify the definition of the objective function.
- The choice of the NSGA-II algorithm is somewhat surprising. NSGA-II is designed for multi-objective optimization, and its core mechanisms are unnecessary or inefficient for a single-objective problem. Since you appear to have a single objective function, why not use a standard genetic algorithm (GA) or particle swarm optimization (PSO)?-- Section 4
- You repeatedly state that dynamic wake mixing is part of your framework. However, FLORIS is a steady-state, time-averaged engineering model and does not inherently simulate transient or dynamic wake evolution. Am I missing something here? How exactly is FLORIS used to evaluate the effects of dynamic wake mixing?-- Conclusion
- “Overall, these findings demonstrate that coordinating multiple control strategies, rather than applying them in isolation,” — do you mean “multiple control objectives” instead of “control strategies”?Citation: https://doi.org/10.5194/wes-2025-240-RC2
Model code and software
Code for an integrated control framework for floating offshore wind farms Yue Niu and Ryozo Nagamune https://drive.google.com/drive/folders/1h6ebwumd06PdEuX38KJgqEwQK9Lhss3e?usp=sharing
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 285 | 72 | 22 | 379 | 16 | 15 |
- HTML: 285
- PDF: 72
- XML: 22
- Total: 379
- BibTeX: 16
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments:
The authors present a new wind farm control approach for offshore applications that combines turbines repositioning, wake steering, wake mixing, and power curtailment. They provide an introduction, a description of their methodology, and finally, the results of their study. The idea behind the work aligns well with current wind energy research on control, making it relevant.
At the same time, several important methodological details are currently missing or underdeveloped. Because the study relies heavily on FLORIS, the paper would benefit from a description of the chosen velocity, deflection, and turbulence models, along with the parameters used and the motivation for selecting them. These aspects have a strong influence on the results, and clarifying them would greatly strengthen the robustness and reproducibility of the findings.
Similarly, the implementation of wake mixing is not fully specified. For example, if the Helix model introduced recently in FLORIS is used, details such as the chosen amplitude and the rationale behind it would be valuable. The interaction between wake steering and curtailment could also be discussed more explicitly, especially given the complex aerodynamic behavior associated with simultaneous misalignment and changes in thrust coefficient.
Moreover, there is no mention of how curtailment is performed, which is rather crucial because it determines how the thrust coefficient changes.
It is also not clear whether the authors are combining wake mixing and wake steering simultaneously. Although this could be possible in FLORIS, there has not been a validation of the results yet. The effects on the wake do not simply sum up, and this would likely result in a critical overprediction of power.
Another point concerns the ambient conditions used in the simulations, such as turbulence intensity and vertical shear. These are key inputs to FLORIS and strongly influence wake recovery, so specifying them would improve transparency.
Overall, I believe that the results may not be reliable without a clear justification for the modeling choices.
To sum up, I believe the paper addresses an interesting and important topic. With a more complete description of the methodology and modeling assumptions, the results would become more reliable and easier for readers to interpret and reproduce.
If the authors are able to incorporate the suggested methodological clarifications, it could also be beneficial to expand the Results section. At present, the analysis feels somewhat limited, and given that the authors have developed a promising control framework, there is an opportunity to demonstrate its capabilities more extensively.
Specific comments:
- Line 52: the references for wake mixing could be updated to include more recent work.
- Line 56: a short discussion of how wake mixing can be implemented, with examples from the literature, would be helpful
- Line 59: a reference is needed to support the statement that the methods are cooperative rather than competitive
- Line 59: it would be good to clarify that wake mixing is not combined with wake steering in the cited work
- Line 109: the controller assumes knowledge of hub-height wind speed. Since this is difficult to measure in practice, adding a brief discussion would strengthen the paper
- Line 126: "Notably, power maximization can be regarded as a special case of pwoer regulation by srtting P_ref to the rated farm output", this sentence could be misleading, since the rated farm output typically refers to the greedy case, not to the power boosting case.
- Section 3: This section includes many repetitions of things that have been said already. It should be shortened in my opinion. For example:
- Line 142: "A central challenge in wind farm operation is the presence of wake effects" this has been already said.
- Lines 143–146, 161–163, 176–177, 184–186: These points appear to have been stated previously.
- Line 150: it is not clear what Beta is. In line 188, there is a brief mention that this is the blade pitch angle. This is then the Helix approach, but there is no mention, description, or reference to it.
- Line 181: reference missing: "Although less effective than repositioning (CIT) ... "
- Line 191: it seems like the authors are combining wake mixing with wake steering. This seems to push the capabilities of FLORIS a bit beyond its limits. As far as I know, there has not been any validation of this combination yet, and the effects do not really add up.
- Line 209: introducing weights in the cost function is sometimes inevitable, but it introduces important choices. How were these weights set and why? Was the cost function normalized somehow? The power tracking error and the velocity error have substantial differences in order of magnitude.
- Section 3.2.1: More detail is needed on how turbines are curtailed. The method strongly affects the thrust coefficient, so describing the approach is important
- Line 242: since a genetic algorithm is used with a large number of scenarios and variables, a short note on convergence criteria or computational considerations would be insightful
- Line 262: how was power maximization achieved with the cost function in Eq. 2? This seems to apply only to power regulation.
- Line 265: was a wind rose considered here? What turbulence intensity was sed?
- Figure 9: showing local inflow velocities would be informative, since they also appear in the cost function.
- Figure 9: the histograms refer to wake mixing, wake steering and integrated control. It would be interesting to see also the effect of repositioning alone.
- Line 306: the power regulation case was run at a wind speed of 18 m/s. Since this is above rated for most turbines, the controller behavior may be less interesting there. Running it in region II might give more insight. A brief explanation of the choice would help.
- Line 340: I encourage the authors to consider open-access sharing platforms, as this greatly supports reproducibility.
Technical comments:
- Line 126: reference missing: "... or ancillary service needs (CIT)."
- Figure 8 is difficult to read in black/white
- Figures 10a, 10b, 11a, 11b, 12a, 12b, 13a, 13b: axis labels should be added.
- In figures 10c, 11c, 12c, 13c, it is not entirely clear what the histograms represent; a short explanation in the caption would help.