Tradeoffs On Improving WRRF Effluent Water Quality With Advanced Analytics

Summary: Understanding the synergistic relationship between wastewater treatment and AWT can help inform the design, modification, and operation of these facilities to ensure a more reliable, economic and safe alternative potable water supply. The goal of this work was to utilize advanced analytics to identify the potential linkages between WRRF parameters and AWT performance. Learning Objectives: Attendees will understand the WRRF parameters that appear to influence AWT performance. Attendees will also gain insight into the benefits of advanced analytics.
Introduction: Traditionally, a water resource reclamation facility's (WRRF) main focus is to meet effluent discharge requirements. For utilities considering implementation of advanced water treatment (AWT) for potable reuse, understanding the synergistic relationship between wastewater treatment-both liquid and solid streams-and AWT can help inform the design, modification, and operation of these facilities to meet the water quality requirements necessary to ensure the safe reuse of water as reliably and efficiently as possible This work was performed as part of the larger collaborative project WRF 4833. The key goal of this portion of the project was to document how performance at a WRRF can impact AWT processes and demonstrate the synergistic relationship between WRRF and AWT through the presentation of multiple case studies. These impacts may be process specific, related to operation and maintenance (O&M), water quality impacts that increase health risks (either acute or chronic) or impact treatability (health/treatability), or a combination of O&M and health/treatability impacts. Identifying the cause and effects of WRRF on AWT performance is intended to help address the challenges often encountered when implementing advanced treatment including: - Significant diversity in the process configurations and operating strategies employed at WRRFs. - Treatment goals at WRRFs that are sufficiently distinct from AWT goals. - Conventional monitoring at WRRFs can be insufficient to quantify impacts on AWT. - Intrinsic variability in wastewater influent, environmental conditions, and biological processes can prevent elucidation of cause and effect at the AWT.
Methods: Five utilities across the United States with varying treatment configurations were selected for the analysis, the treatment configurations are summarized in Table 1. Multi-linear regression analysis in PythonTM and supervised learning in BayesiaLab was performed to identify the potential linkages between WRRF parameters and AWT performance.
Results: Multi-linear regression analysis and supervised learning were able to develop predictive relationships between WRRF parameters and AWT performance with varying accuracy (0.5 < R2 < 0.98), example predictive plots are shown in Figure 1. The results of the advanced analytics evaluation identified several key WRRF parameters that appear to influence AWT performance. In general, the analysis demonstrated the following: - GAC effluent TKN is influenced by primary clarifier surface overflow rate, number of clarifiers in service, and digester hydraulic retention time - GAC effluent TP is influenced by primary clarifier SOR, number of clarifiers in service, and alum use - GAC effluent COD is influenced by primary solids flow and bioreactor dissolved oxygen (DO) - Microfiltration flux is influenced by solids retention time and clarifier scum flow - Microfiltration transmembrane pressure is influenced by TSS load, BOD removal, and COD removal - Reverse osmosis flux is influenced by MLSS, sludge volume index, and solids retention time - Reverse osmosis recovery is influenced by clarifier scum flow, WAS flow and yield, TSS removal, and solids production Additional work is needed to further understand the explanatory nature of the data. Machine learning tools such as unsupervised learning and semi-supervised learning may help to identify clusters of WRRF parameters that influence AWT performance and improve the ability to detect anomalies in treatment performance. Nevertheless, the results of this study provide a preliminary understanding of the interrelationship among WRRF parameters and key AWT performance.