By Kumar and Parks from MIT. They propose a hyperelastic material model for graphene based in invariants. I don't remember the details of the paper, but it seems that you could use this to code a UMAT or a UANISOHYPER iin abaqus.
I don't understand your question regarding a 2D or a 3D shell. You can use either, I'm not too familiar with 3D shells. If I understand correctly you are referring to finite element implementations that consider the shell as degenerate solid as opposed to a mid-surface description? I'm definitely not familiar with the degenerate solid approach. Is it most commonly used for thick shells? For thin shells definitely the midsurface description is applicable. You also need a thickness though.
You can use beam elements to model the bonds between the atoms in a graphene sheet :-)
Creating a beam element is an easy task which needs two points. If you have the coordinates of all atoms using some programming language, you can write a code to connect the points.
I also engage in the graphene sheet simulation. But I got some troubles due to convergence. Have you solved your problem about the graphene simulation?