Optimisation of Water Quality Management in Drinking Water Distribution Network System using Computational Methods

Abstract

Ensuring the delivery of safe drinking water through complex distribution networks remains a paramount public health challenge globally. This study presents a comprehensive computational framework for optimising water quality management in Drinking Water Distribution Networks (DWDN) through the development and validation of a digital twin integrated with advanced machine learning methodologies. The research establishes a novel digital twin framework that simulates both hydraulic and water quality dynamics using EPANET, coupled with a Physics-Informed Neural Network (PINN) surrogate model for enhanced predictive accuracy. A robust data preprocessing pipeline was developed to assimilate flow, pressure, chlorine residual, pH, and turbidity data from operational distribution systems. The integrated framework demonstrated exceptional predictive performance, achieving a model accuracy of 95.2% for chlorine residual prediction and 94.7% for hydraulic state estimation. The comparative analysis revealed that the proposed PINN-based surrogate model outperformed conventional EPANET simulations by 18.3% in computational efficiency while maintaining superior accuracy in water quality parameter forecasting. The digital twin framework enables real-time anomaly detection and proactive management of water quality degradation events. This research contributes a scalable, computationally efficient methodology for water utilities to enhance operational decision-making, ensuring regulatory compliance and safeguarding public health through optimised distribution network management.