This paper considers the control of dissolved oxygen concentration in a wastewater treatment plant. Since the system is nonlinear and multivariable in nature, a complex control system should be postulated. However, under certain conditions it is feasible to decompose the control problem into separate multi-input single output sub-systems. In this case, the control loops of substrate and dissolved oxygen concentrations in the bioreactor can be independently handled and thus, more effectively regulated. A non-linear predictive control algorithm is proposed to satisfy the process demands. Robustness of operation and ease of constraint incorporation make the long range predictive control philosophy particularly attractive. Dissolved oxygen is regulated by a series of aerators with fixed speed motors. It is the sequence of motor on/off switches which must be determined in order to control the dissolved oxygen level. Furthermore, several constraints must be satisfied in order to avoid mechanical or technical problems. The non-linear predictive control algorithm has been developed to predict the series of motor switches which will result in tight dissolved oxygen regulation whilst also satisfying these constraints. The results obtained show that the proposed controller may be a good alternative to the present control philosophy.