The application of artificial neural networks for the optimization of coagulant dosage

Filtration is the final physical barrier preventing the passage of microbial pathogens into public drinking water. Proper pre-treatment via coagulation is essential for maintaining good particle removal during filtration. To improve filter performance at the Elgin Area WTP, artificial neural network (ANN) models were applied to optimize pre-filtration processes in terms of settled water turbidity and alum dosage. ANNs were successfully developed to predict future settled water turbidity based on seasonal raw water variables and chemical dosages, with correlation (R2) values ranging from 0.63 to 0.79. Additionally, inverse-process ANNs were developed to predict the optimal alum dosage required to achieve desired settled water turbidity, with correlation (R2) values ranging from 0.78 to 0.89.