For a finite system, it can be simulated by a supercell with sufficiently large lattice constant or by an isolated system with open boundary condition for the electrostatic potential. In the former case, only one k point (e.g., usually Gamma point) is enough . There is no k point sampling in the latter case.
" If I calculate a very large system(such as 100*100 unit cells in transverse plane) ,whose supercell is too large to calculate directly,can I just build a bulk model and sample the k points by 100*100 to estimate the result?"
You cannot.
As already pointed out, the size of k point sampling does not reflect the simulated system size.