Overall quality:

Interesting work, the case study is relevant and the proposed procedure has been well described. The text is quite clear and easy to follow. 

However, I see two important points to improve:

• The authors should elaborate on why the frequency interval from 1/2 h^-1 to 1/3 h^-1 has been selected for the optimization instead of larger periods (e.g. 1-2 days) or a wider range of frequencies (e.g. from daily to 1/2 h^-1). This choice should be carefully motivated since it is at the article’s core. One could argue that the 2-3h period falls right within the spectral gap of typical wind variations at a site and therefore relatively low variability is expected in that range. Valid arguments justifying the frequency range selection should be presented already in the introduction (e.g. around lines 31 or 59) and then renewed in the discussion sections (e.g. around lines 157 and 237).  
• The authors should present at least a hypothesis for why the optimization provides a limited benefit compared to an even distribution of the capacity. Moreover, it would be nice to see some comments about what would be expected if the same methodology was applied to different case studies.

Specific comments:

• Title: “… to minimize the overall power fluctuations …” – maybe it would be better to rephrase it a bit, e.g. with something like: “… minimize the power fluctuation at selected frequencies …”
• Abstract: “The focus is mainly on the smoothing effect in highest resolvable frequencies.” – it would be nice to add a brief explanation for why these frequencies are relevant.
• Abstract: “decrease the 1-3 hourly fluctuations considerably” – it would be better to be more quantitative.
• Line 94 – The authors should explain why that height was selected. Is it the turbine’s hub height?
• Line 96 – Please motivate the choice of this turbine and give at least some minimum specifics like the hub height and the rated power.
• Line 108 – Please indicate the overall duration of the signal and the chunks.
• Line 210 – It would be nice to link this to the findings of previous studies.
• Line 269 – The paragraph on future work is a bit superficial. Possibly it should be extended or at least better argumented.

Technical corrections:

• The labels or at least the captions of the PSD plots should specify what quantity is considered and its physical dimensions (or if normalized it should be mentioned). Also, I think that adding gridlines to the plots would help their interpretability.
• I could find several typos and a few small mistakes in the use of English. I will only list a few here, but please check the manuscript thoroughly before submitting the revised version.
• Abstract: “5th and 95th percentiles” – specify “of the hourly step-change functions”.
• Line 37 – extent
• Line 42 – recent
• Line 58 – geography
• Line 60 – is available
• Line 162 – constraint
• Line 210 – pronounced
• Line 270 - planned

This paper is addressing the minimization of the overall power fluctuations for different wind farm portfolios. In general, the proposed method has the potential for publication in WES. I have the following comments:

• The 5th and 95th percentiles of the step change function of wind power time series are used as statistical variability metrics in this paper, which from my point of view represent the range of power fluctuations. It could be more descriptive if the authors were also looking at mean and standard deviation.

• This paper should be restructured to improve its readability. The Section Introduction needs to improve. Please follow this sequence: problem definition and motivation for research in this field, literature review, and the main contributions of this research. Subsection "A note on ignoring the wind farm smoothing effect" could move to the introduction as the paper assumptions. The Section “Data” could merge with the Section "Result". Optimization theory in Subsection “Optimization of wind farm capacities” is better to move to Section "Method".

• There are language errors and typos, e.g., constrain instead of constraint in pg 8 line 162.

• Illustrated PSD in Fig. 5 is a little confusing. What are the time step and the time interval for the PSD analysis performed in this figure?

• Could the authors bring more details into the mathematical presentation of the optimization objective function represented by eq.2?  The PSD of which function is going to be minimized in the specified frequency range.

• Have the authors tested different frequency ranges, and why is the frequency range (2h)-1 (3h)-1 chosen for the optimization?