Although, only in recent years, northern European countries started to install large offshore wind farms, it is expected that by 2020, several dozens of far and large offshore wind farms (FLOWFs) will be built. These FLOWFs will be constituted of a considerable amount of wind turbines (WTs) packed together, leading to an energy density increase. However, due to shadowing effects between WTs, power production is reduced, resulting in a revenues decrease. Therefore, when FLOWFs are considered, wake losses reduction is an important optimization goal. This work presents a modular approach to optimize the energy yield of FLOWFs through an evolutionary algorithm. The method consists of a modular strategy where the site wind rose information is used in different steps, which accelerates the calculation of the wake losses. The main contribution of this paper is the use of surrogate models to optimize the layout of offshore wind farms. Although, the surrogates models do not make use of the entire wind information set, they preserve the main problem trend. At the end, the results obtained are tested for their sensitivity regarding the wind data and the turbine locations.