WCW24

Coagulation chemistry is a complex process impacted by several variables and is regarded as “the single most important factor affecting treatment plant performance” (AWWA M37). Machine learning is a subset of artificial intelligence that has the potential to ‘learn' complex chemical processes. The City of Lethbridge Water Treatment Plant has used machine learning modelling of the coagulation process as an operational support tool since February 2023. The tool provides recommendations to operators in real time based on raw water characteristics and treatment system conditions to achieve multiparameter optimization. In the first year of implementation this tool has helped operational staff optimize treatment during both stable and rapidly changing raw water conditions, resulting in a reduction in coagulant and polymer usage by approximately 21% while exceeding treatment objectives. Over the last year assessments have been completed at several other conventional water treatment plants in Western Canada to assess the transferability of the technology. The assessments have included WTPs with varying source waters, coagulants, clarification technologies, and levels of instrumentation. This presentation will provide an overview of coagulation, machine learning, and the results of the WTP evaluations. Specific challenges related to data gathering, exploratory data analysis, and modelling will be highlighted. Potential challenges and benefits of the technology will be discussed, and the overall results of the assessments will be shared.