I want to do MD simulation on a protein model to refine its structure. My protein structure have both folded and intrinsically disordered regions. What force field should i choose for this goal?
There is no best or worst forcefield, as different ff has been developed with specific goal in mind. I have seen a wide variety of ff in literature, so in most cases it is just personal preference.
However, Amber and Charmm are suitable for this purpose. AMBER99sb-ildn is better and inherently available in gromacs.
For further reading, you can consult the following paper (and other similar papers):
These days both Amber and CHARMM have more dedicated FFs to treat disordered regions. The "bare" Amber99SB, even with ILDN, is not guaranteed to produce meaningful results for IDPs/IDRs.
DE Shaw's group published their 'disp' correction that was recently made available for Gromacs by Paul Robustelli: https://github.com/paulrobustelli/Force-Fields (you might need to switch to their version of TIP4P to get optimal performance, though).
In parallel, CHARMM's version 36m was specifically designed to improve the treatment of IDRs: https://www.charmm.org/charmm/showcase/featured-research/charmm36m-an-improved-force-field-for-folded-and-intrinsically-disordered-proteins/
You will find ample of publications on simulations for protein having both structured and disordered states defining CHARMM or AMBER as a better choice.