We study the exclusion of salt from charged porous media (Donnan effect), by using a coarse-grained approach. The porous medium is a lamellar system, namely a Montmorillonite clay, in contact with a reservoir, which contains an electrolyte solution. We develop a specific coarse-graining strategy to investigate the structural properties of this system. Molecular simulations are used to calibrate a mesoscopic model of the clay micropore in equilibrium with a reservoir. Brownian Dynamics simulations are then used to predict the structure of ions in the pore and the amount of NaCl salt entering the pore as a function of the pore size (the distance L between clay surfaces) and of the electrolyte concentration in the reservoir. These results are also compared to the predictions of a Density Functional Theory, which takes into account the excluded volumes of ions. We show that the calibration of the mesoscopic model is a key point and has a strong influence on the result. We observe that the salt exclusion increases when kappaL decreases (where kappa is the inverse of the Debye length) and that this effect is modulated by the correlations between ions. Two different regimes are revealed. At low concentrations in the reservoir, we observe a regime controlled by electrostatics: the Coulomb attraction between ions increases the amount of salt in the interlayer space. On the opposite, at high concentrations in the reservoir, the excluded volume effect dominates.