The application of neural networks to model a full-scale industrial wastewater treatment plant has been studied. A hybrid first-principles neural network model has been elaborated and successfully tested. In the proposed model, the trained neural network represents the unknown kinetics of biological decomposition of pollutants. This net has been trained with experimental data obtained from a long-term monitoring of an existing industrial wastewater treatment plant. For the considered plant, sufficiently good accuracy of the dynamic behavior predictions has been obtained.