Hi Nico,
Here is a wiki page about PRE with a few literature links. Briefly, with PRE decrease of T2 relaxation time is measured introduced by a covalently attached paramagnetic label, such as a nitroxide-based radical. radical.
When label is at a fixed position on a target molecule, increase of relaxation rate R2 (or decrease of T2) of certain nuclear spin resonances is correlated with a distance between the label and the nucleus. The shorter is distance - the stronger is the relaxation enhancement and the enhancement will be proportional to inverse sixth power of distance.
With this method distances up to ~25A can be accessed, whereas NOE shows only short distance contacts of <5-6A.
As for how this can be applied to your cellulose studies - not so clear. Can you clarify in your question - what problem are you trying to solve? Also - what is the molecular weight of the cellulose species? Keep in mind that with flexible molecules distance restraints obtained by PRE or NOE may not be meaningful, because they will be averaged and heavily biased towards shorter ones (due to the R^-6 dependence).