Experimental Measured Data Driven Diagnosing and Improving Biological Phosphorus Removal at the Great Lakes Water Authority (GLWA) Water Resource Recovery Facility (WRRF)

Background and Objectives
The Great Lakes Water Authority (GLWA) Water Resource Recovery Facility (WRRF) serves 3.1 million residences in Southeast Michigan. Treating a combined sewage inflow up to 3,560,000 m3/day (940 million gallons per day, MGD) it is one of the largest treatment facilities in the world. The secondary treatment process consists of four limited-nitrifying High Purity Oxygen (HPO) activated sludge bioreactors, consisting of eight cells in two bioreactors and ten cells in the other two bioreactors, followed by 25 circular secondary clarifiers. Phosphorus removal is achieved by a combination of ferric chloride addition to influent wastewater prior to primary treatment along with coincidental biological phosphorus removal (bio-P) occurring in the HPO system. Influent wastewater characteristics are highly variable, likely due to sedimentation and biological activity occurring in the large volume and high residence time of the combined sewer system.
This paper summarizes the results of over four years of work: (1) characterizing variations in the influent wastewater characteristics, (2) documenting the unique bio-P performance of the system resulting principally from fermentation occurring in the secondary clarifiers (a form of side-stream fermentation), and (3) consolidating these results in a plant-wide model characterizing chemical (ferric chloride) and biological phosphorus removal along with impacts of solids handling recycles. These results should be of interest to those working to achieve effective phosphorus removal along with those dealing with extreme wet weather flow conditions.
Methodology
Experimental work consisted of extensive analysis of 8-2/3 years of liquid stream operating data (from January 2013 to August 31, 2021), coupled with influent wastewater characterization, full-scale bioreactor profiles, batch reactor testing, and jar testing. Process modeling was conducted using SUMO21 (Dynamita), calibrated as depicted in Figure 1. Figure 2 is the developed model from this study to reflect actual operating conditions for modeling historical performance (i.e. calibration). Batch reactor tests conducted in the late summer of 2018, early winter of 2020, and early winter of 2021 assisted with model calibration. The tests were performed to simulate the anaerobic and aerobic periods occurring in the full-scale HPO bioreactors and within the secondary clarifier sludge blanket. Appropriate quantities of return activated sludge (RAS) or concentrated mixed liquor were combined with primary effluent, then sampled for measuring ortho-phosphorus, soluble COD (sCOD), and oxygen uptake rate (OUR). An initial half-hour anaerobic period was followed by two-hour aerobic period per one cycle of batch tests. In some cases, a further two-hour anaerobic period was added to simulate residence time within the secondary clarifier to measure bio-P's performance in different seasonal conditions. In 2021, another set of batch reactor was operated under strict anaerobic conditions for 72 hours to measure phosphorus release rate and possible phosphorus release amounts. Wastewater characteristics over the entire wastewater treatment region were separately analyzed for 5-days biological oxygen demand (BOD5), total chemical oxygen demand (COD), sCOD, total phosphorus (TP), total SP (TSP), total suspended solids (TSS), volatile suspended solids (VSS), and influent volatile fatty acid (VFA) concentration. A year-long intensive sampling phase evaluated particulate, colloidal, and dissolved biodegradable, and non-biodegradable organic matter.
Results
Influent fractionation is a crucial component of calibrating the model and predicting bio-P performance from modeling because biodegradable organic amounts (including soluble particulate forms) largely affect microbes' activities in the system, and it is generally assumed the fraction of biodegradable organic matter from influent VSS as 0.625. Although this assumption returns a reasonable estimation, it cannot capture the nature of fluctuating influent wastewater, resulting in deviations in model prediction. An interesting correlation was observed from the dataset (Figure 3), showing a robust correlation between influent BOD/VSS concentration and its removal ratio over primary clarifiers, dBOD/dVSS (here dBOD and dVSS refer to the removed fractions, respectively, of the two components). This leads to estimate fraction of biodegradable particulate organic matter from influent VSS, allowing reasonable approximation of daily varying wastewater characteristics for simulation. The modeling results are displayed in Figure 4 where MLVSS was used as an indicator for the calibration performance over the input mass balance. The bio-P performance was calibrated using batch test results (Figure 5), and Figure 6 compares the results predicted by the calibrated model. We find that influent VFA concentrations were insufficient to support bio-P activity, and the initial unaerated zones were not large enough to either supply or produce the VFA's needed for PAO growth. However, we found that sludge fermentation in the secondary clarifiers did provide sufficient VFA's for PAO growth. Analysis of the model results identified three general factors affecting bio-P performance: (1) For the GLWA WRRF, the data suggests that an SRT of 1.8 days or shorter at a wastewater temperature of about 10 °C leads to PAO washout, and subsequently to elevated effluent TSP. (2) During warm periods, even small amounts of nitrite and nitrate generated by nitrification adversely impact bio-P because they act as an alternative electron acceptor in the secondary clarifier sludge blanket, resulting in reduced fermentation and VFA production for PAO uptake and phosphorus release, which suppresses the PAO population. (3) A sufficient sludge blanket depth must be maintained to achieve sufficient fermentation and VFA production for PAO growth. Conclusion Influent characteristics are utilized to estimate time varying influent particulate organic matter fraction, which improves model prediction performance. The calibrated modeling results suggested operational protocols for improving bio-P performance, these include: (1) seasonal adjustment of SRT (range of 2.0 to 2.5 days) in response to wastewater temperature variations to avoid PAO washout during low-temperature conditions and to minimize nitrification during warmer periods, (2) maintaining a secondary clarifier sludge blanket of 3.3 feet. Minor physical modifications are also suggested, including the addition of mixers to the bioreactor's initial zones and the diversion of pure oxygen feed around these aeration deck stages to improve bio-P performance via the establishment of deeper fermentation conditions. The fermentation also could be increased by directing RAS to the first stage of the aeration deck and using step feed capabilities to direct influent to the second stage. The results of suggestions show the improved bio-P indicated in Figure 7.
Discussion
The effect of cessation of ferric chloride addition to the influent wastewater was simulated by adjusting PE, TP, and TSP concentrations to account for historical chemical phosphorus removal. Figure 8 presents estimated effluent TP level in comparison to revised influent TP concentration assuming ferric chloride is not added to the influent wastewater. The results indicate that plant effluent TP discharge limits can be met by developing these suggested conditions, suggesting that ferric chloride addition could be reduced or eliminated while sustaining required performance.