General Comments

The manuscript presents a novel control strategy, COFLEX, for large flexible wind turbines, addressing a critical gap in optimizing turbine performance considering structural flexibility. This study is highly relevant as the increasing scale of wind turbines introduces significant challenges in structural dynamics and control. The integration of a set point optimization framework (COFLEXOpt) with a feedforward-feedback control scheme represents a substantial improvement over traditional tip-speed ratio-based methods.

The manuscript effectively builds upon existing methodologies and tools for optimization, estimation, and control. While none of the individual modules (optimization framework, wind speed estimator, or controller) are entirely new, the parameterization of dimensionless coefficients in three dimensions instead of two is a noteworthy adaptation. These adaptations are well-discussed, seamlessly integrated, and tested comprehensively through simulations. The paper is clear, well-structured, and includes detailed methodologies and results.

Specific Comments

• Design of Wind Turbines: The manuscript could benefit from a brief discussion of passive design techniques such as pre-coning/pre-bending of blades to mitigate flexibility effects and tailored bend-twist coupling for passive load alleviation.
• Objective Function:  The motivation for solely minimizing the torque and power coefficient in the steady-state optimization is unclear. It is recommended to (a) describe the optimization framework in a more general form and (b) include a discussion on other meaningful objectives to demonstrate versatility of the proposed framework. The use of a single objective function across the entire operating region is intriguing as objectives are considered to vary between full and partial load operation.
• Model Limitations: A discussion on robustness of the proposed control approach is recommended, including potential discrepancies between HAWC2 simulations and real-world turbine dynamics. How do modeling uncertainties and measurement errors affect performance?
• Wind Speed Estimator: The wind speed estimator described in Section 5.1 is tested using a Kw2 control law. Please clarify which power coefficient is used in the Kw2 law and discuss if/how the selection of the feedback gain K impacts the estimator’s performance.
• Saturation Schedules: A clearer motivation is needed why saturation limits are computed solving the “reduced” optimization described in Equation (18).  Could alternative formulations also be used?

Technical Corrections

• Line 165: The mention of "direct drive" appears misplaced and should be revised for clarity.
• Equation 2 needs to be revised.
• It is "HAWCStab2" instead of "HAWC2STAB".

This is a very clearly written manuscript that discusses a new control strategy that combines feedforward torque and pitch control, using optimized control commands scheduled using a wind speed estimator, with feedback control to ensure the rotor speed setpoint is tracked. The paper does a nice job of showing the importance of optimizing pitch and rotor speed setpoints in the partial load region as a function of wind speed, using an aeroelastic turbine model with rotor flexibility, rather than assuming a single optimal tip speed ratio and blade pitch. This is especially important for highly flexible rotors where blade deformations due to thrust affect the aerodynamic properties of the rotor throughout the partial load region.

This paper builds on previous work examining how turbine models including blade flexibility can be used to optimize control set points while adhering to design constraints (and the advantages over traditional control laws). Specifically, this paper extends previous ideas by presenting a closed-loop control strategy to implement the intended set points using a wind speed-estimator-based combined feedforward/feedback controller with smooth setpoint switching between the partial load and full load regions. The incorporation of blade tip displacement constraints in the set point optimizations is another important contribution.

I don’t have any major concerns with the paper, but there are several areas where I believe corrections or clarifications are needed or some additional analyses would help provide more value. 

Comments:

1. Pg. 4, ln. 91: Given the similarity of this work to Pusch et al. 2023, please explain the differences between that study and the research in this paper here.
2. Pg. 8, ln. 207: Can you add the resolution of the rotor speed, wind speed, and pitch angles that make up the 27,000 points? Also, can you clarify if wind shear is included in the inflow?
3. Pg. 10, ln. 241: "These likely unrealistic large torsional deformations…" Please explain why you believe these are unrealistic deformations.
4. Section 4, 1st paragraph: Minor point. It would be nice to mention what Section 4.2 covers in this introduction.
5. Page 11, ln. 262: "As a consequence, the rated wind speed and operating regions were predefined." This doesn’t appear to be true. In Pusch et al. 2023 (https://onlinelibrary.wiley.com/doi/10.1002/we.2879), Section 3.1 states that "rated generator torque and  speed are not pre-defined herein and are subject to optimization as well… the ratio of rated generator torque and speed is determined at the smallest wind speed where a given value of rated generator power is reached." Can you clarify in more detail how your approach differs from this previous study?
6. Pg. 13, ln. 310: "due to the small contribution given by the torque coefficient term" You explain that the weighting term w_1 for the torque coefficient should be small, but how did you choose the specific value? What value was finally used?
7. Pg. 16, ln. 343: "where the blades pitch in to relieve thrust force" Should this be pitch "out"? Larger blade angles (from pitching out) would lead to lower thrust generally.
8. Eq. 10: I believe the inertia term "J" should be in the denominator of both of the fractions on the right hand side of the first line.
9. Pg. 19, ln. 399: "able to estimate the wind speed at a steady state with a significantly smaller error" Can you discuss what might cause the small error in the wind speed estimates for the Flex. 2 case? Are there additional degrees of freedom in the simulation that aren’t in the wind speed estimator model?
10.  Pg. 21, ln. 423: To match the description of the feedback pitch command in this sentence, you could state that when e_omega > 0, Delta Q_g,FB should similarly be negative to accelerate rotor speed.
11. Eq. 17: Is Beta_max also 30 degrees in this case?
12. Pg. 23, ln. 447: "producing stable operating points for the wind turbine in the full load region" Please explain how this choice of j(V) produces operating points that are "stable" and how this differs from the strategy used by Abbas et al. 2022. How would the stability compare to other simpler choices of j(V), such as setting it equal to the pitch angle at the rated wind speed?
13. Pg. 23, ln. 455: "In the partial load region, Delta omega_bias_s = 0 and Delta omega_bias < 0" Is there a sign error somewhere in Eq. 16 or 19? Otherwise I think Delta omega_bias would only be negative in the partial load region if the gain K_bias_1 is negative (I'm assuming you intend for the gains to be positive values).
14. Pg. 24, ln. 464: How did you design this low-pass filter? What cut-off frequency was used?
15. Fig. 15: Can you explain why there is an underestimation bias in the estimated wind speed?
16. Fig. 15b: There is a considerable high-frequency component in the blade pitch feedforward signal (and the torque setpoint too) stemming from the high frequencies in the estimated wind speed. Can you please discuss where this comes from, and is it problematic? The oscillations in the pitch angle could potentially increase damage to the pitch actuators. What improvements could be made to reduce the high-frequency component of the estimated wind speed?
17. Fig. 15d: e_omega is consistently negative across the 20 seconds of the simulation. This suggests that the combined pitch and torque control strategy is regulating rotor speed poorly. Can you discuss this? I'm also surprised that the difference between e_omega and e^prime_omega is so small. Given that the difference is tiny compared to the magnitude of the rotor speed error, how does this meaningfully impact the setpoint switching?
18. Fig. 15: Minor point, but in the first sentence of the caption it might be clearer to describe the variables in the order they're shown in the subplots.
19. Pg. 28, ln. 527: "a discrepancy in the steady-state blade deflection calculation for HAWC2 and HAWCStab2": Could you simply use HAWC2 for the steady-state calculations?
20. Section 6.2: It would strengthen the results of the paper to compare the controller performance in turbulent wind conditions to the performance of the simpler reference controller. For example, although the steady state results show improvements compared to the reference controller, how do the power and tip displacement compare in more realistic turbulent conditions between the novel controller design and the reference controller?
21. Section 6.2: Please mention what amount of wind shear was included in the turbulent simulations.
22. Pg. 28, ln. 555: "This consistent, positive bias… " Could this positive bias be caused by the presence of wind shear? For example, shear might lead to higher turbine power than predicted by a simple rotor average of wind speed used for comparison.
23. Pg. 29, ln. 563: "the expected constraint on the median value of the OoP tip displacement is satisfied with a deviation of less than 1%". For OoP tip displacement, I would think the maximum value would be more important than the median within a wind speed bin (since even one tower strike would cause damage). Is the median value a relevant way to judge the tip displacement here?
24. Section 7: This is impressive work, but it would be interesting to briefly summarize ideas for improvements to the control strategy for future work. For example, could the controller be designed to better handle different amounts of wind shear? How could the controller be combined with IPC to better reduce maximum tip displacement? Could the wind speed estimator be improved to reduce the high-frequency ripple in the estimates? Are there ideas for reducing the bias in the wind speed estimator?