Aeration control based on a neural network in a biological aerated filter for simultaneous removal of ammonia and manganese

Abstract

This study was conducted to monitor and control aeration by means of an online Neural Network (NN) of a Biological Aerated Filter (BAF). The BAF is an advanced drinking water treatment system equipped with Dissolved Oxygen (DO), oxidation-reduction potential, pH, ammonia and nitrate sensors. The main function of the BAF is to treat contaminated water by simultaneously reducing the levels of ammonia and manganese to below permit limits. Aeration was supplied to the BAF and controlled by a neural network. Real-time data was accurately predicted by the NN with errors below 5% for all sensors. The bending point was apparently created in DO neural network data when the simultaneous ammonia and manganese removals were below limits. The NN program detected the bending point and immediately stopped the aeration of the BAF. Hence, NN can optimize the aeration requirement and system performance, shorten time demand and reduce consumption of manpower and electricity.