General comments

Dear Amr and co-authors, I have reviewed your article and I would like to provide my feedback on it.

The article describes a wave basin experiment investigating the effect of a wave-feedforward control strategy on the global response of a floating wind turbine scale model. The experiment is carried out by means of the software-in-the-loop technology: the floater, mooring system and the environment are modelled physically at small-scale, whereas the wind turbine rotor, its aerodynamics loads, and the wind turbine controller are emulated with a numerical model and actuators.

I think the topic of the article is interesting for readers of this journal. Wave feedforward might have the potential to improve operation of future floating wind turbines. Moreover, the wave feedforward control strategy and its testing with software-in-the-loop are multidisciplinary problems, thus the findings of this article can be beneficial for all the community working on (floating) offshore wind turbines. The article is generally well written, the methodology is sound and its results are clear. For all these reasons, I believe the article deserves to be considered for publication on the Wind Energy Science journal.

I have some comments for the authors that I hope will improve the article. Feel free to continue the discussion if my comments are not clear.

The experiment is carried out with the software-in-the-loop (SIL) methodology, which is presented in the article. With SIL you are coupling the physical and numerical domains by means of measurements and actuators. In my experience, this coupling can be affected by issues such as delays and it’s not perfectly “transparent” (i.e., it perturbs the wind turbine response compared to a case where all subdomains are physical). In this sense, do you see any effect of the SIL technology itself on the results of the experiment? Maybe you can add some short comments on that in Sect. 2.3 or in the Results section.

The feedforward controller is derived based on a model of the floating wind turbine that is obtained by means of system identification applied to a QBlade model. What did you model in QBlade: the full-scale wind turbine or the scaled system. I assume the scaled floating platform is not a perfect model of the full-scale one (e.g., it has different mass distribution, mooring response, ...). In case you modeled the full-scale (ideal) wind turbine, do you expect that this can affect the results of the experiment? Also in this case, you may add a comment on that in Sect. 2.3 or in the Results section.

After the first part of the experiment, it was decided to continue testing on one of the two FF controllers only. I think this choice is not sufficiently motivated (or the motivation is not very clear to me). The FF for platform pitch requires 20% more blade pitch actuation compared to baseline. Is this above some kind of limits of the wind turbine (e.g., maximum pitch rate)? Or did you decide to stop testing the FF for platform pitch because it gave less benefits compared to the FF for generator power but requiring higher control effort? From Fig. 10, it seems the FF for platform pitch is more effective with respect to its control objective (~-20% of platform pitch variation) than the FF for generator power (<-5%) so I think it's not clear why you decided to continue with the first controller only. If I understood the results correctly, and if you agree with my comments, I think you should also say that you tested both controllers in a baseline scenario, then you decided to explore the performance of wave FF control in a broader range of conditions and, to do so, you focused on one on the two controllers.

Specific comments are reported below, and technical corrections are in the attached document.

Specific comments

• 12-13. “It was concluded that […] response than waves.”. I suggest removing this sentence and summarizing in a quantitative way the results of the experiment.
• 33-35. “As a result, […] to disturbances.”. This sentence should be rephrased. The pitch controller is slower and has limited authority against disturbances of wind and waves.
• 77-78. “Whether the scaled-model […] of the FOWT.”. I suggest rephrasing this sentence. It is not a matter of complexity but which domain you want to reproduce with higher accuracy (wind in the wind tunnel, water in the wave basin).
• 86-89. “When it comes […] (Al, 2020).”. I would skip these sentences. As you say, the FF controller was not tested.
• 132-134. “This accelerometer […] thrust forces.”. I would remove this sentence because here it is not clear why you have to remove the inertia forces. Explain it in the SIL section.
• 138-140. “The numerical simulation […] on the elements of the blades.”. Explain this in the SIL section.
• Section 2.3. Even if it seems trivial, I suggest adding a sentence in this section to explain that in the SIL approach the wind turbine rotor is not scaled physically but it is replaced by force actuators that emulate the rotor loads.
• Caption of Figure 4. I would remind the reader that the numerical model emulates the wind turbine controller and aerodynamic response.
• Figure 5. If I understand correctly you use an algorithm to predict wave loads and this is the input of the wave FF controller (not the wave elevation). Can you add one figure next to this one to showcase the wave force prediction?
• 202-203. “takes a structure […] second-order system”. How did you choose this structure?
• Figure 7-8. I suggest merging figure 7 and figure 8 in a single figure with two subfigures.
• 207-208. Same as comment for lines 202-203.
• “the frequency band of interest enclosed by the vertical lines”. How did you define this frequency band?
• 216-218. Rephrase these sentences. I suggest saying that you test both controllers in one condition, then you proved that one is more effective than the other, hence you carry out all other experiments with the first controller only.
• 222-225. You can remove these sentences if you rephrase the first part of the introduction of Sect. 4 according to the previous comment. I think that what is explained in lines 222-225 must be clarified before Table 3.
• “We begin with illustrating the two different control objectives”. You are studying the performance of the two controllers and not illustrating their objective (this was done before).
• 228-229. “At the end […] such control.”. This part is not clear. At the end of the comparison, you understand that one control objective can be reached with reasonable control effort, but not the other. Then you carry out experiments to understand the potential of one of the two controllers considering a wider range of operating conditions. It should be also mentioned that you compare the two controllers to the baseline feedback controller to assess their performance.
• 237-238. “which indicates […] wave signals”. It's not clear what you mean. I think you should say that when you add the FF controller to the FB controller, the low-frequency content of the signal remains the same and you have additional blade pitching at the wave frequencies.
• 244-245. “The PSDs […] other objective.”. I don't understand this sentence. I think you mean that reduction of power fluctuations with the FF controller for power is obtained with less blade pitch actuation that reduction of platform pitch motion with the FF controller for platform motion.
• “the effect of the feedforward controller for both control objectives.”. I think you should add “because the two controllers are expected to decrease the variance of power and platform pitch counteracting the effect of the wave disturbance”.
• Once you define one symbol, it’s not necessary to define it again. I would avoid defining symbols more than once to improve readability.
• 255-256. “and we even […] reduction FF control”. This sentence is not clear but I don’t have a suggestion to improve it. Please double check it.
• “the thrust force varies”. Is this the aerodynamic thrust or the total thrust force (aerodynamic + structural)?
• 272-274. “a huge actuation effort […] for the rest of this study”. See the comment in the General comments.
• 301-302. “the peak-to-peak value of the signal is decreasing as the turbulence intensity increases”. I'm not sure about this comment. With TI=0%, the peak-to-peak is about 2°, with TI=13.8%, the peak-to-peak is about 8°.
• 229-330. “as there is no clear […] for every DOF separately”. These two sentences are not clear to me.
• 346-347. “we can not […] the generator power signal”. Do you have an explanation for that? If yes, maybe you can add a short comment.
• Figure 19. If Fig. 19 does not add value to the discussion you can remove it and just explain by words that blade pitching is the same in all cases you analyzed.
• Through the result section, we see the response of the platform DOFs is not coherent, and they react differently to changes in the environmental conditions. Do you have an explanation for this behavior? In case, you may add some short comments in the results and in the conclusion where you discuss the platform response.
• 351-355. “This is because … of the wave fluctuations”. Can you rephrase these sentences? I think they are not clear.
• Section 4.5. all wind speeds are above the rated wind speed. Can you add a short comment to explain why you did not consider below rated wind speeds?
• Data availability. I think it is mandatory to have this section. Is there any data/model made available for readers (also by contacting you)?

Technical corrections

See attached document.

The paper presents a wave feedforward control for a floating wind turbine in a wave tank. The experiment is based on software in the loop, where the spar is physical while the aerodynamic is operated in the aeroelastic code QBlade. The feedforward control design is based on the loop shaping technique, where the transfer functions are formed based on frequency response in order to achieve the control objectives: minimise the rotor speed and platform pitch. The paper is well-written. Many environmental conditions were considered e.g. wave height, turbulence intensity and wave spreading. The topic is highly relevant and valuable for the readers of Wind Energy Science, as control experiments in scaled floating wind turbines are rare in the literature. Therefore, I would recommend the manuscript for publication after minor clarification of my following comments.

1. Page 6. “where kp and ki are the proportional and integral gains respectively, which were properly tuned using the loop-shaping technique.” Could the authors elaborate more? How did the baseline controller get tuned? Did you reduce the gain crossover frequency to overcome the negative damping problem compared to onshore cases?
2. Page 11. “In order to obtain the TFs; Gωg,β(s) and Gθp,β(s), a chirp signal, logarithmically distributed over the experiment’s duration,” As I understood, these transfer functions are not physical in the experiment. What is the reason you didn’t compute them directly using aeroelastic code instead of using the system identification method?
3. Also, did you update your feedforward controller design in different operating conditions, as the frequency responses are different?
4. Figure 6. I understand that the feedforward controller maps the wave-induced moment to blade pitch. Could you elaborate on how the wave LiDAR information is translated into the wave-induced moment?
5. Figure 6. Did you use the Constant Torque strategy in the experiment? Can you justify it?
6. Figure 7 & 8. The phase response of your design (red) looks quite different to the actual frequency response (blue). A phase difference would cause a time delay in the FF control action, derived from the disturbance, which might affect the disturbance rejection performance. Could you comment on this?