Back-propagation neural network adaptive control of a continuous wastewater treatment process

Abstract

Wastewater treatment processes and technology have been investigated for several decades and have almost been completed up to date. In this study, a chemical method was applied to treat the wastewater. Instead of real wastewater, benzoic acid and water were mixed as the wastewater since different concentrations of dissolved benzoic acid could result in different chemical oxygen demands (COD). Hydrogen peroxide and ferrous chloride were both added to treat the wastewater in order to meet the standards of 1998 environmental regulation in Taiwan. pH was found to be a major factor affecting the coagulation condition of the suspended paticles during the treatment process. Back-propagation neural network was applied, and the purpose of the control was to provide the minimum amount of reagents to reach the required COD. The pump rates for adding hydrogen peroxide and ferrous chloride were controlled. The neural network was of a time-delayed mode, and its structure was properly determined, with the only output node being the predicted H2O2. The concentration of the added reagents was compared as well.