Development and Comparison of Water Quality Network Model and Data Analytics Model for Monochloramine Decay Prediction

Abstract

The conventional drinking water treatment process involves disinfecting water at the final stage of treatment to ensure water is microbiologically safe at customer taps. Monochloramine is a popular disinfectant used in many water distribution systems (WDSs) worldwide. Understanding the factors that impact monochloramine decay in the WDS is critical for maintaining disinfection at the customer tap. While monochloramine residue moves through a WDS, it decays via several pathways including chemical, microbiological, and wall decay processes. The decay profile in these pathways is often site-specific and depends on various factors including treated water characteristics. In a water quality network model, the decay of a chemical species is often modelled using two parameters that represent bulk and wall decay kinetics. Typical bulk decay characteristics of monochloramine for a specific WDS can be easily established in the laboratory using grab sample tests, while in a real situation, wall decay is difficult to quantify. In this study, we compared two different approaches to model monochloramine decay in a WDS. In the first approach, the wall decay parameter was quantified using a parameter optimisation technique with monochloramine concentrations at different network locations simulated using a water quality network model. In the second approach, a data analytics model was developed using a machine learning algorithm. For both approaches, the model predicted monochloramine concentrations closely matched the observed data. Our study suggests that the data analytics model has a relatively higher accuracy in predicting monochloramine residual concentrations in a WDS.

Keywords

• monochloramine decay modelling
• EPANET
• water quality data analytics model
• support vector regression