Use of artificial neural networks for predicting optimal alum doses and treated water quality parameters

Abstract

Coagulation is an important component of water treatment. Determination of optimal coagulant doses is vital, as insufficient dosing will result in undesirable treated water quality. On the other hand, doses that are too high can result in high cost and health problems related to high levels of residual aluminium (if alum is used as the coagulant). Traditionally, jar tests are used to determine the optimum coagulant dose. However, this is expensive and time-consuming and does not enable responses to changes in raw water quality in real time. Modelling can be used to overcome these limitations. In this paper, artificial neural network models are used to model alum dosing of southern Australian surface waters. The performance of the models is found to be very good, with correlation ( R 2) values ranging from 0.90 to 0.98 for the process models predicting treated water turbidity, colour and ultraviolet absorbance at a wavelength of 254 nm (UVA-254). An R 2 value of 0.94 is obtained for the process inverse model used to predict optimum alum doses. Two simulation tools, Sim TTP and Sim WT, are developed to enable operators to obtain optimum alum doses easily. In addition, the simulation tools enable alum dosing rates to be controlled automatically in real time.