Global solar irradiances at ground level are modelled over Belgium using latest version of WRF-ARW regional climate model (RCM). The model set-up used has a resolution of 5 kilometres. The boundary layer scheme chosen is the Mellor-Yamada-Nakanishi-Niino (MYNN) 2.5 scheme with the Turbulent Kinetic Energy (TKE) closure proposed by Canuto et al., (2008) and Kitamura (2010). In this scheme, the modification of some parameters allows to change the determinant mixing length (surface layer, planet boundary layer, top of boundary layer/entrainment) which then modifies heat and moistures fluxes produced by turbulent mixing. Such modifications have significant influences on modelled cloudiness and therefore on modelled global solar irradiance incoming at the surface. The present study proposes a sensitivity analysis of the different parameters that influence the mixing length ('alp1' to 'alp5') and the TKE diffusion ('Sqfac') in order to find the most suitable constant values of these parameters for the modelling of cloudiness over Belgium. Results of different simulations are compared with global solar irradiance measurements performed by the Centre Spatial de Liège at Sart-Tilman in 2013 and 2014. Firsts results show that the dry bias frequently found when using WRF-ARW with standard set-ups can be greatly reduced thanks to an increased modelled cloudiness. The quantitative and qualitative effects of these modifications over cloudiness are also analysed by displaying 2D representation of modelled clouds over Sart-Tilman and confronting them with on-site observations.