Radial basis function artificial neural network able to accurately predict disinfection by-product levels in tap water: Taking haloacetic acids as a case study

Abstract

Control of risks caused by disinfection by-products (DBPs) requires pre-knowledge of their levels in drinking water. In this study, a radial basis function (RBF) artificial neural network (ANN) was proposed to predict the concentrations of haloacetic acids (HAAs, one dominant class of DBPs) in actual distribution systems. To train and verify the RBF ANN, a total of 64 samples taken from a typical region (Jinhua region) in China were characterized in terms of water characteristics (dissolved organic carbon (DOC), ultraviolet absorbance at 254 nm (UVA254), NO2−-N level, NH4+-N level, Br− and pH), temperature and the prevalent HAAs concentrations. Compared with multiple linear/log linear regression (MLR) models, predictions done by RBF ANNs showed rather higher regression coefficients and accuracies, indicating the high capability of RBF ANNs to depict complicated and non-linear relationships between HAAs formation and various factors. Meanwhile, it was found that, predictions of HAAs formation done by RBF ANNs were efficient and allowed to further improve the prediction accuracy. This is the first study to systematically explore feasibility of RBF ANNs in prediction of DBPs. Accurate predictions by RBF ANNs provided great potential application of DBPs monitoring in actual distribution system.