Advancing Full-Scale Partial Denitrification-Anammox (PdNA) Filters: Design and Operation Guidelines

Introduction
Partial denitrification (PdN) coupled with anammox (AnAOB), referred to as PdNA, has the potential to considerably increase the capacity of existing nitrogen removal systems in upstream nitrification as a lower concentration of ammonium needs to be oxidized. By balancing the ammonium oxidation with NOx, the aerobic SRT, aeration energy and alkalinity needs would decrease. This will potentially allow for more effective use of influent carbon in upstream anaerobic and anoxic zones. Moreover, the external carbon used for denitrification could be lowered by up to 80%1. Polishing filters are widely used for the simultaneous removal of TSS and nitrate. Applying PdNA in polishing filters would allow reaching stringent effluent limits in an energy and resource efficient manner. In the quest to successfully advance the PdNA technology in deep-bed granular media filters, 6 pilot-scale studies, 1 full-scale study in the winter and 1 full-scale study in the summer were conducted by our group. This paper discusses the experiences gained from those studies to evaluate the reliability and robustness of this promising technology to accelerate the full-scale implementation.
Results and discussion
PdNA filters do not require a different media type Fast adoption of PdNA filters is more likely if existing filters can be retrofitted. The conventional medium – silica sand – was compared to expanded clay, which previously demonstrated higher AnAOB growth and retention at bench-scale testing2. At pilot-scale, expanded clay didn't show any significant benefits over the silica sand tested (Table 1), suggesting easy retrofits of conventional denitrification filters into PdNA filters. This was also demonstrated in York River Treatment Plant. York River deep-bed denitrification sand filters were transitioned into PdNA sand filters in 2018, making it the first full-scale mainstream anammox implementation. Conditions favorable to AnAOB growth were provided by bypassing some ammonium to the filters and maintaining a nitrate residual concentration. As AnAOB built up, a more optimized aeration control, ammonium/NOx (AvN) control, was used in the activated sludge process to enhance performance.
The carbon source will depend on the unit cost The performance of the conventionally used methanol was compared to glycerol, proven efficient in other PdNA studies3-5. The higher performance of the glycerol (Table 1) wasn't significant enough to compensate for the difference in yields and costs, therefore the carbon choice for PdNA filters will depend on the unit cost. Carbon savings compared to full denitrification were estimated at 58% and 65% for methanol and glycerol. The full-scale PdNA filters, operating with methanol, showed lower PdN efficiencies and PdNA rates compared to the pilot-scale filters due to its lower loading rates (Table 1). However, the transition from full-denitrification filters to PdNA filters resulted in 80±30% methanol savings. Thus, conventional methanol-based filters could easily be retrofitted into PdNA filters.
Phosphorus requirements A substantial decrease in PdN efficiency, PdNA rates, and runtime was observed when phosphorus demand was lower than 0.03-0.04 g OP/g TIN in the glycerol filter. Decrease in nitrogen removal was also observed in conventional filters under phosphorus limited conditions6, However, such trends were not observed in the methanol-based PdNA filter, which showed similar demand as conventional filters6. This indicated that glycerol-based PdNA filters might require a higher need in phosphorus for microorganisms to sustain their growth.
Control optimization requirement A process-control is required for carbon addition in PdNA filters similar to conventional filters. In these studies, feedforward/feedback and feedback controls were used. Theoretically, lower COD/N ratios are required in PdNA to halt the nitrate reduction at nitrite. For instance, at the full-scale, the COD/N ratio decreased from 4.4 to 1.8 g/g, which ultimately lowered the methanol requirements significantly. However, a nitrate residual needs to be maintained to achieve a high PdN efficiency. Therefore, automated and effective control of carbon dose is required to provide substrate for AnAOB. In addition, a well-balanced NH4+/NOx stream is necessary to meet the stoichiometry requirements for PdNA filters. AvN control in the upstream activated sludge process is required to provide ammonium and NOx to maximize AnAOB contribution given a certain PdN efficiency, thus impacting the filters effluent concentrations. The results in these studies (Figure 1) agreed with conceptual7,8 and practical9 studies that showed the importance of AvN in one-stage PdNA systems without additional polishing zones to reach low effluent limits. Optimization in activated sludge process aeration control is therefore essential to meet effluent limits; and the AvN setpoint is determined by PdN efficiency, total nitrogen and ammonium effluent targets. Microbial communities Candidatus Brocadia, found in other PdNA systems10-13 , was the only AnAOB genus found in these studies (Table 1). Denitrifiers, previously found in PdNA systems14-16, were also identified (Table 1). As methanol was used in the full-scale filters, methylotrophs were also identified (Table 1).
Capacity and robustness to backwash Maximum AnAOB activity tests under non-limited substrate conditions revealed excess ammonium removal was up to 3 times (pilot-scale) to an order of magnitude (full-scale) higher than the daily operational rates in the summer (Table 1), suggesting excess AnAOB biomass accumulation. Although the activity decreased in the winter, the maximum activity was still 1.5 (pilot-scale) to 5 (full-scale) times higher than the operation, indicating that AnAOB are not limited in PdNA filters, and significantly more nitrogen removal can be achieved. The pilot filters were subjected to different backwash procedures to evaluate which conditions resulted in AnAOB loss. No loss was observed to the point where daily performance was impacted when the backwash frequency was increased. Only when the air scour and backwash times increased to unrealistically high values on the order of 5.5 hours, a decrease in maximum AnAOB activity by a factor of 3 was observed but 80% on the incoming TIN was still removed daily. AnAOB activity recovered two days later, suggesting that AnAOB were not washed out, but regrowth of heterotrophs was probably needed to protect AnAOB mass from oxygen or carbon to restore full activity.
Conclusion
The results indicate a clear path toward retrofitting conventional denitrification filters into PdNA filters, and greenfield applications are also recommended. It was shown for the first time that the PdNA technology is applicable in filters under representative loading rates similar to conventional filters. In addition, the conventional sand and methanol have shown increased nitrogen removal rates through the PdNA route. A switch to glycerol could also be feasible only when its unit cost is lower than the methanol. Overall, AnAOB are not the limiting steps in the implementation of PdNA in filters. The success of the PdNA technology is rather dependent on the aeration control in upstream activated sludge process to provide the right AvN ratio to maximize AnAOB contribution given a certain PdN efficiency.