The recent growth of the size of wind farms highlights the need for a deeper understanding of the mesoscale phenomena in a stably stratified atmosphere, such as atmospheric gravity waves, as the effect on the power generation can be significant. This thesis is a numerical study of the effect of wind farm layout on atmospheric gravity wave excitation and the resulting feedback on wind farm performance. Wind farms in this study have varying power density, streamwise or spanwise turbine spacing, aspect ratio, hub height, rotor diameter, shape, or orientation with respect to the freestream, and are situated in a conventionally neutral boundary layer in offshore conditions. Moreover, the wind farms can be horizontally or vertically staggered. To investigate the atmospheric gravity wave excitation, the effect of wind farm layout on the Froude number and inversion Froude number governing the internal and interfacial waves respectively is studied using several high-fidelity Large Eddy Simulations. Then, AGW wavelength and wind farm efficiency are parametrically studied using a fast reduced-order model. Specifically, the non-local efficiency is considered, which is a measure of the global blockage effect induced by the atmospheric gravity waves. It is found that the length scale used in the Froude number must be adjusted to the wind farm, and is dependent on the turbine spacing and farm shape. The inversion Froude number is based on the phase speed of the interfacial waves. It is suggested that the phase speed must be based on shallow-water theory and deep-water theory for small and large turbine spacings respectively. In other words, for small turbine spacings the wind farm acts as an entity, while for large turbine spacings, the farm acts as a collection of individual turbines. Finally, the streamwise and spanwise turbine spacing (and consequently the power density), and the aspect ratio primarily govern the non-local efficiency.