Graphene features unique electronic, thermal, and mechanical properties, and the flexibility and strong attraction between graphene layers promotes the formation of self-folded nanostructures. Here we study the self-folding of mono- and multilayer graphene sheets, utilizing a coarse-grained hierarchical multiscale model derived directly from atomistic simulation. Our model, developed by enforcing assertion of energy conservation, enables the simulation of graphene folding across a range of length scales from nanometers to micrometers. Through theoretical and simulation analysis we show that the critical self-folded length is πC/γ, where C and γ are the bending stiffness per unit length and the surface energy per unit length.