Dear Mostafa,
Yes, I agree with your least-squares regression approach.
So, does your LES simulation include a wind turbine or does the LES simulation only consider the atmospheric boundary layer? If a wind turbine is modeled within your LES, I agree picking points in the ambient wind say 2D upstream of the rotor makes sense. If the wind turbine is not modeled within the LES, then I would pick points at the rotor plane.
Are you cubing the wind speed in your spatial averaging to derive some sort of power-equivalent average wind speed? I agree that there are many points that you could derive the turbulence intensity from and I’m not sure whether there is a best practice on the approach taken. I would first judge how different the turbulence intensity is across the domain. If you want to limit the number of points considered, I would use points at the hub height.
Best regards,