Real-time Optimization of Peracetic Acid Disinfection at Metro Wastewater Reclamation District Using Artificial Neural Networks

Peracetic acid (PAA) is a particularly difficult disinfectant to model at full-scale and requires expensive analyzers for real-time measurements; requiring wastewater treatment plants (WWTP) to operate with a large safety factor to ensure disinfection performance. The Robert W. Hite Treatment Facility (RWHTF), owned and operated by Metro Wastewater Reclamation District in Denver, Colorado, saw a significant discrepancy between pilot-scale and full-scale PAA disinfection performance resulting in overdosing to ensure complete disinfection. In this work, CT-based PAA dosing strategies are compared to optimize dosing for RWHTF. Methods include conventional integrated CT with and without an online PAA analyzer, and CT predicted using data-driven modeling methods. To model at full-scale and respond real-time water quality changes, artificial neural networks (ANN) were trained to predict PAA at various locations throughout a 3-pass serpentine disinfection basin and integrated CT using online and laboratory data from upstream treatment processes. The ANN model predicted CT within ± 14 mg/L∙min without an online analyzer and ± 12 mg/L∙min with an online analyzer which is equivalent to predicting CT 47% and 30% more accurately than the conventional CT control, with and without an online analyzer respectively.