A series of experiments were undertaken in order to understand and predict the dosage of powdered activated carbon required to remove taste and odour compounds in an Australian drinking water treatment plant. Competitive effects with organic matter removal by aluminium sulphate during coagulation were also quantified. Data on raw and finished water quality following jar tests, as well as chemical dosages and treatment performance, were statistically analysed, and a data-driven prediction model was developed. The developed powdered activated carbon dosage prediction model can be used by the plant operators for rapid dosage assessment and can increase the preparedness of the plant to sudden taste and odour events. It was also found that total organic carbon levels and properties greatly affect the ability of powdered activated carbon to remove taste and odour compounds; on the other hand, total organic carbon removal is not affected by high taste and odour levels, since these were still much lower than organic carbon concentrations.
Keywords: Data-driven modelling; Powdered activated carbon; Taste and odour; Water treatment optimization.
Copyright © 2018 Elsevier Ltd. All rights reserved.