The Water Research Foundation (WRF) is conducting a study to explore how real-time machine learning (ML) can optimize dosing of disinfection chemicals in wastewater utilities, aiming to improve compliance and reduce operational costs. This initiative, termed WRF project 5385, titled Machine Learning Meets Disinfection: Full-Scale Wastewater Chloramination Control for Compliance and Cost Savings, is led by Carollo Engineers, alongside Ekster & Associates and the Los Angeles County Sanitation Districts (LACSD). The project will implement an ML-based disinfection control framework at a water reclamation plant (WRP) and develop practical guidelines for other utilities.
Building on insights from WRF project 5148, which successfully applied ML in aeration control at LACSD’s Pomona Water Reclamation Plant (POWRP), this new project aims to extend those methodologies to disinfection processes. The previous project achieved significant reductions in blower energy usage and nitrate levels, as well as savings in chloramination costs.
Many current disinfection systems rely on traditional feedback controls that do not anticipate rapid changes in flow or water quality, leading to conservative chemical dosing and increased costs. The new project will assess if ML-driven predictive control can better manage these variables by adjusting dosing in real-time while ensuring compliance. Over 24 months, the team will develop predictive models, integrate control logic, deploy necessary instrumentation, and validate performance.
Dr. Natalie Beach, the principal investigator for the study and the east region wastewater lead at Carollo, stated:
“When compliance is critical, conservative dosing is logical. This project aims to evaluate whether machine learning can more effectively anticipate changing conditions and safely modify disinfection dosing in real-time, through a phased approach that allows operators to gain confidence before moving to direct control. Ultimately, a key objective of this initiative is to ascertain if machine learning can achieve similar enhancements in operational efficiencies and cost savings as our research team accomplished when applying machine learning to optimize and automate the biological nutrient removal process at the Pomona Water Reclamation Plant.”
A final report will share findings and guidance for utilities interested in advanced disinfection control strategies.
Source: Carollo Engineers
