In this study, the author investigates real-time optimal control of wind farms using large-eddy simulations. Overall, this manuscript is of good quality and novelty. However, I still have some doubts that require further clarification in the current manuscript by the authors.

1. In section 2.3, the authors discussed the turbine modelling used in this study and proposed a look-up table approach as an ad-hoc solution to fix the low-resolution issue (~2 grid points across the rotor diameters). However, the main concern of this reviewer is that the look-up table in Appendix A is constructed based on uniform inflow, neglecting the effects of turbulent inflow and vertical wind shear that are critical to the power of wind turbines in a wind farm). As a quick check, this reviewer would like to ask the authors to examine the correcting factor M in a turbulent inflow generated by the precursor simulation and check if it significantly differs from the uniform inflow results.
2. In section 4.3.1, the authors discussed the computational cost. However, since the simulation in this study required a precursor simulation to drive the flow, this reviewer would like to ask the authors to discuss the cost of precursor simulation in this analysis.

The paper discusses the implementation and testing of wind farm control strategies (combination of wake steering by yawing and reduction of axial induction) synthesized by means of an LES-based control model.  Specifically, the authors aim to investigate the impact of the receding horizon parameters, control update frequency, and grid resolution on the computational time. The final objective of their efforts is to enable a real-time implementation of the proposed control methodology. All results are obtained in simulations, with a fine-grid LES model used as plant emulator.

The paper is interesting, describes a novel approach, and is well-written. I have a single major comment.

The gains of case 4 (circa 50%) are much higher than the gains with steady yaw control (circa 10%, as also reported in Sood and Meyers, 2022). The authors claim that "Simulations (not shown here) suggest that yaw control only (i.e. disabling induction control) does not entail a significant performance reduction.", which implies that the large difference between the gains of case 4 and of Sood and Meyers is not due to enabling the reduction of induction control in case 4. The authors also claim that case 4 is characterized by quasi-static yawing behavior. For this case, the yaw angles, indeed, do not exhibit major variations, but rather minorly oscillate around the quasi-static optimal yaw angles. It is reasonable to expect the quasi-static optimal yaw angles to be close to the optimal yaw angles of the steady yaw control of Sood and Meyers (in this regard, it would be helpful to include the optimal yaw angles of Sood and Meyers in Figure 13). , I, therefore, wonder what the reason for the very big differences between the achieved gains of case 4 and of the steady yaw control. Please argue on this.

Minor comments:

• Figure 9, rrror in caption: PDE, not L-BFGS-B
• Line 516, typo in "investigated"