A precise dosing system based on a coagulation reaction model for water treatment

Coagulant dosing in water treatment plants typically relies on operators' subjective judgments, supplemented by water quality data and experimental records. This approach cannot effectively adapt to dynamic water quality changes, leading to compromised water quality and increased costs. Most solutions to this problem involve machine learning models that learn from human judgments. However, poor control of the coagulant dosage, often due to interoperator variability, leads to unstable settled water quality. Operators also often overestimate dosages to ensure compliance with purification standards, and these empirical methods typically overlook the impact of flocculation mixing speed, leading to excess coagulant use and higher energy consumption. This study proposes a precise dosing system based on a coagulation reaction model, which considers the physical and chemical relationships between water quality and operational parameters. Using jar tests, we collected real-world coagulation data across different raw water qualities and operational conditions, developing a machine learning model to predict settled water quality accurately. The trained model was integrated into the proposed system to estimate the optimal parameters for achieving the required water quality standards at minimal cost and carbon emissions. On-site evaluations demonstrated a 33.7 % reduction in coagulant dosage, an 81.4 % decrease in flocculation energy consumption, and a 20.6 % reduction in carbon emissions, outperforming existing empirical models that only achieve up to 10 % coagulant reduction and neglect mixing speed's influence. Our proposed system significantly enhances operational efficiency, cost-effectiveness, and sustainability.