drift during MD Forcite simulation

Hi I have bummer. For the first time in MS my system is drifting. I want to calculate self-diffusion of N2 in PEG polymer net. I have eq. the structure in NVT and running NVE from current velocities. My system is drifting. I thought that MS automatically cancels the non zero momentum but this time no. When I calculate MSD of N2 instead of line i got a parabolic dependency. Any idea? I attach msd, status file of sim and initial structure file