This paper shows how using in-reactor on-line monitoring of OUR in the aerobic reactors of an A2/O (anaerobic, anoxic, oxic) nutrient removal system, the implemented control system is able to improve the nitrogen removal of the plant. The control system adapts the operation to different load conditions, without the need for additional information provided by other analysers. In particular, these OUR measurements and on-line information available in the distributed control system (temperature, biomass concentration and flows) are used for the estimation of the COD load in the influent, and automatically applied in the adaptation of the operational mode of the nutrient removal process. The estimation is realised using Artificial Neural Networks models trained with available data from the system. The operational changes consist in the modification of the total aerated volume in the pilot plant as a function of the load estimation. With the implemented control system based on COD load estimation, an increase in nitrogen removal of 10% was obtained with respect to a system working with fixed aeration volume, and using a suitable amount of energy in each case.