Application of computational intelligence for on-line control of a sequencing batch reactor (SBR) at Morrinsville sewage treatment plant

Abstract

Morrinsville Sewage Treatment Plant has recently been upgraded to an Extended Aeration SBR. The plant needs to comply with stringent discharge requirements despite the variations in organic and hydraulic load caused by tradewaste discharges and stormwater infiltration. Effluent data from a nearby dairy factory is transmitted to the treatment plant by radio and processed by a back propagation neural network trained to correlate the data with the corresponding BOD. BOD oxidation, nitrification and denitrification rate constants are estimated by fuzzy systems as function of temperature and MLVSS. Output data generated by the model are used to assist control of SBR cycle duration, sludge wasting, and temporary storage of excessive load in a lagoon. The model does not pretend to provide an accurate description of the process, nor a fully optimised control system, but rather a common-sense approach to the very challenging operating conditions. This is a plant receiving a low level of supervision and it is expected that the control system will improve process performance and compliance with discharge requirements.