Development and performance of a machine learning-based tool that predicts influent flow to a WRF 72 h in advance and integrates with existing wet weather nutrient management protocols.

Influent flow to the 75 mgd Neuse River Resource Recovery Facility (NRRRF) was modeled using machine learning. The trained model can predict hourly flow 72 h in advance. This model was deployed in July 2020, and has been in operation over two and a half years. The model's mean absolute error in training was 2.6 mgd, and mean absolute error has ranged from 10 to 13 mgd in deployment for any point during the wet weather event when predicting 12 h in advance. As a result of this tool, plant staff have optimized the use of their 32 MG wet weather equalization basin, using it approximately 10 times and never exceeding its volume. PRACTITIONER POINTS: A machine learning model was developed to predict influent flow to a WRF 72 h in advance. Selecting the appropriate model, variables, and properly characterizing the system are important considerations in machine learning modeling. This model was developed using free open source software/code (Python) and deployed securely using an automated Cloud-based data pipeline. This tool has been in operation for over 30 months and continues to make accurate predictions. Machine learning combined with subject matter expertise can greatly benefit the water industry.

Enhanced coagulation using a magnetic ion exchange resin.

The objective of this investigation was to examine the effectiveness of a magnetic ion exchange resin (MIEX) to enhance the coagulation of disinfection by-product precursors in nine surface waters, each representing a different element of the USEPA's 3 x 3 enhanced coagulation matrix. The effect of MIEX-pretreatment on the requisite alum dose needed for subsequent coagulation of turbidity was also evaluated. Enhanced coagulation with MIEX was found to be very effective for removing trihalomethane (THM) and haloacetic acid (HAA) precursors from the nine waters examined. THM and HAA formation potential was reduced by more than 60% in all of the waters studied; reductions approaching 90% were seen in the waters with the highest specific ultraviolet absorbance values. The residual total organic carbon concentration, ultraviolet absorbance, and THM and HAA formation potential were all substantially lower as a result of MIEX and alum treatment compared to alum coagulation alone. MIEX pre-treatment also lowered the coagulant demand of each of the waters substantially.