A review of data-driven modelling in drinking water treatment

There are significant opportunities to optimize drinking water treatment and water resource management using data-driven models. Modelling can help define complex system behaviour, such as water quality and environmental systems, giving insight into expected outcomes from changing conditions. Many water treatment models have been developed, such as predicting treated water quality based on coagulant addition or disinfection by-product formation from chlorination, which can be used to better inform operators of optimal treatment parameters to limit risk and reduce cost. Data-driven models, in particular, present an opportunity to learn relationships from patterns in historical data without the need to pre-define mechanisms or variable interactions. Furthermore, models built on currently monitored data are likely easier to implement since they utilize water quality measures that are already in place. However, data-driven approaches have significant challenges, including increased uncertainty in model validity, challenges in interpreting model behaviour and decision logic, and increased likelihood of incorporating biases from training data. This article presents a review of data-driven model applications in drinking water treatment to highlight opportunities related to protecting source water quality, optimizing treatment processes, and interpreting of sensor data. There is a focus on identifying approaches and algorithms best suited for specific applications and the interpretability of trained models. Successful implementation of data-driven models in critical systems, such as water treatment, requires that models be validated, and a model’s decision-making logic can be identified and scrutinized.