Has your system equilibrated? I find that whenever I’ve tried applying force to a carbon sheet to mimic pressure, the carbon sheets’ separation fluctuates wildly – I attribute that to carbon being “too light”. A poorly compressible system will throw out big pressure waves en route to equilibration, especially if started from an improperly small or large initial volume, and a piston with too little density like carbon will fluctuate in response, often too much for a typical MD box to handle.
(Or maybe I’ve just been doing it wrong all this while?)
In your position I would start simple: run a “constant volume” simulation by completely not integrating the wall particles; track the density profile of the water across the system using some fix ave/chunk; if the water density right in the middle of the polymer-free zone is correct (1 g/cc) and the water density permeating into the polymer is stable over time, then you at least have a baseline for attempting barostatted simulations. You can even estimate the pressure on the walls using compute group/group between (for example) the left wall and all non-wall atoms; the average total wall force will be the effective pressure, times cross-sectional area.
Alternatively, use fix langevin with a very high friction coefficient to damp the graphene walls’ vibrations, instead of whatever thermostat you currently use. How high? High enough to get the “correct answer”. What’s the correct answer? I wouldn’t know, but the constant volume simulations would give you at least some clue.