In order to effectively control the influence of Concentration Polarization (CP) during the nanofiltration separating wastewater process, this study applied parameters characterization of membrane flux attenuation coefficient (mwt) and the Back-propagation (BP) neural network algorithm to simulate the development rules of CP and membrane pollution, set up CP BP Model of Nanofiltration Separation, based on the tested data of NF90. The correlation coefficient between simulation and test of the simulation BP model was over 0.99, with the absoluteness error below 1.5%. According to the model’s prediction, the separation effect of nanofiltration technology become attenuate with running time increasing in nanofiltration separating wastewater process. mwt start raised obviously within first 0.5h in operation and stay stable after 1h. It was advised to appropriately maintain u>0.2m/s for NF90 membrane effectively controlling mwt<0.1.