In this paper, we describe the results of research aimed to evaluate the possibility of using a neural network (NN) model for predicting biological nitrogen and phosphorus removal processes in activated sludge, utilising oxidation reduction potential (ORP) and pH as NN inputs. Based on N and P concentrations predictions obtained via the NN, a strategy for controlling sequencing batch reactors (SBRs) phases duration, optimising pollutants removal and saving energy, is proposed. The NN model allowed us to reproduce the concentration trends (change in slope, or process end), with satisfactory accuracy. The NN results were generally in good agreement with the experimental data. These results demonstrated that NN models can be used as "soft on-line sensors" for controlling biological processes in SBRs. By monitoring ORP and pH, it is possible to recognise the N and P concentrations during different SBRs phases and, consequently, to identify the end of the biological nutrient removal processes. This information can then be used to design control systems.