Not all force fields are parametrized for modelling every material. To model carbohydrates you need to use force fields which are capable of properly modelling these type of substances. You might need to look into GLYCAM06, GROMOS, or CHARMM to get good parametrizations for cellulose, as well as some other carbohydrates.
GLYCAM parameters can be found in Amber. You need to use "leap" to load the parameters, and you can then see the list of residues for cellulose. You basically need ROH, 4GB, and 0GB residues to build up a cellulose model. Next, you can use Acpype or other scripts, for conversion of the topologies to GROMACS format. (an alternative is to use carbohydrate builder on the GLYCAM webpage)
CHARMM and GROMOS on the other hand have been already written for GROMACS use. Looking into the .rtp and .atp files will give you what you need to know about residues.
Note that, to use the latter two, you need to find the force filed files first, and include them in the "top" directory, where all other default force fields in GROMACS are available, for GROMACS to recognize the new force fields. (for these two, just search GROMOS/CHARMM for carbohydrates and you will find what you need)