Partial-Denitrification/Anammox As A Path To Infrastructure and Operational Savings For WWRFs Facing Stringent Nitrogen Limits

In the context of stringent effluent nitrogen limits, there appears to be no better alternative to Partial-denitrification Anammox (PdNA) for continuous flow BNR process upgrades or new construction. Anaerobic ammonia oxidation reduces the need for aerobic ammonia oxidation, drastically reducing required aerobic SRTs; this allows existing plants to increase capacity or dedicate existing aerobic volumes to anaerobic or anoxic volume and allows for designs with smaller nitrification safety factors for new construction. PdNA also reduces aeration requirements, alkalinity addition requirements, supplemental COD requirements, reducing O&M costs and driving sustainability. PdNA does not require anammox seeding from a sidestream process. Polishing PdNA has been demonstrated in full-scale deep bed sand filters for over 3 years (1) and has been demonstrated at pilot scales in MBBR, IFAS, and granular configurations. Startup of these processes appears to take 3 to 4 months (2) which is concordant with sidestream anammox process startup times. Finally, PdNA processes can achieve low effluent nitrogen concentrations (2,3,4). The aforementioned benefits of PdNA stem primarily from the incorporation of anammox into mainstream processes, even though Partial Nitritation Anammox (PNA) theoretically offers slightly improved benefits over PdNA. However, the challenge of maintaining NOB outselection in full-scale continuous flow processes has yet to be overcome (5),and PdNA processes configurations are compatible with PNA if mainstream NOB outselection becomes a reality. This paper explores these claims about PdNA through modelling two BNR processes: an A2O plant with a denitrification filter converted to a polishing PdNA filter and a 5-stage Bardenpho where integrated PdNA (IFAS) is installed in the second anoxic zone. The models were developed in SUMO using the recently developed partial denitrification (PdN) model (6), and the models were compared with existing full scale and pilot data to ensure reliability. Models in this study were built in SUMO with influent characterization shown in Table 1. Process configurations are shown in Figure 1 and Table 2. Both models incorporated suspended growth and biofilm modeling and used default kinetic parameter sets from the SUMO PdNA Model with minor changes. Both models were run to develop a baseline for each process. To incorporate PdNA, the A2O model was modified with AvN control to achieve a target ammonia/NOx ratio in the aeration effluent, which allowed the downstream methanol filter to be transitioned to PdNA via reduced methanol dosing. The 5-Stage Bardenpho process was also modified with AvN control to achieve a target ammonia/NOx ratio in the first aerobic zone effluent, and a biofilm component was added to the secondary anoxic zone to simulate IFAS which allowed PdNA via reduced methanol dosing.
Real-world examples referenced in the paper include sensor data, sample data, and batch test data from: the York River Treatment Plant (YRTP) full-scale process, a pilot-scale PdNA MBBR process, a pilot-scale PdNA IFAS process, a pilot-scale PdNA filter pilot, and pilot tests of fixed-film PdNA media. As anticipated from stoichiometry and previous full-scale and pilot experience, the models demonstrated the benefits of implementing polishing or integrated PdNA. A summary of these baseline and PdNA upgrade model results is given in Table 3. The A2O process baseline required a 10 day SRT to achieve full nitrification and performed as expected for a well designed plant. PdNA was implemented by reducing the aeration tank DO setpoint slightly (2.0 to 1.3) and reducing the total plant SRT (10 to 3.2 days), producing an aeration effluent with an NHx/NOx ratio close to 2.0. Methanol dosing was reduced to drive partial denitrification (PdN efficiency = 65%) in the filters, which resulted in sufficient ammonia and nitrite concentrations for anammox to grow. Startup of the PdNA process in the model required approximately 100 days, which reflects observed PdNA startup times (80-90 days) without seeding (2). The total plant SRT was reduced by 68% while also reducing the alkalinity required by 13% and the methanol required by 52%, demonstrating significant operational savings. Although the OUR in the aerobic reactor was reduced, the aeration requirements did not change significantly, due to changes in alpha factors based on MLSS; and total sludge production for the plant increased slightly, as a reduction in SRT drives higher observed yields. The results of this model aligned well with full-scale results from the YRTP filter performing year-round PdNA (Figure 2): ammonia removal via anammox was in the 2 – 3 mg N/L range and relative reductions in methanol requirements were similar. The 5-stage process baseline was similar to the A2O, requiring a 10 day SRT for full nitrification and achieved acceptable effluent N concentrations. PdNA was implemented by reducing the total plant SRT to 3.4 days, producing an aeration effluent with an NHx/NOx ratio close to 2.0. IFAS (fixed biofilm model) was added to the post anoxic zone and methanol dosing was decreased to drive partial denitrification (PdN efficiency = 58%). This results in sufficient ammonia and nitrite concentrations for anammox to grow, and given the same anammox kinetics were used, the startup time was similar to the A2O PdNA process (~100 days). Again, significant improvements were demonstrated: 66% reduction in SRT, 18% reduction in alkalinity required, and 55% reduction in methanol required. Aeration requirements were essentially unchanged and sludge production increased by 10% due to the large reduction in SRT. These model results aligned well with existing experimental work; the model did show slightly better effluent nitrogen performance than experimental PdNA IFAS has demonstrated (4), likely due to slightly higher PdN efficiency. HRSD is moving forward with construction of integrated PdNA in post anoxic zones at multiple full-scale plants. These results demonstrate not only the benefits of PdNA, but that the knowledge gained thus far about this process can be successfully incorporated into plant-wide models and used to model alternative designs for treatment plant upgrades and new construction. The examples here concentrate most of the gains from PdNA in to reduced SRT requirements, which would lead to increased plant capacity or increased plant efficiency (for plants lacking sufficient anaerobic or anoxic SRT), and still generates significant supplemental COD and alkalinity savings. If capacity is not a primary driver, AvN could be implemented via DO control, reducing aeration requirements and costs significantly and increasing SND, or any combination of reduced aerobic SRT, lower operating DO, transition of aerobic zones to anoxic/anaerobic, etc. Further research with these models will include influent variability (diurnal, wet weather, etc) and more detailed control mechanisms to deal with that variability. Additional modelling work will also focus on sizing and design loadings for these processes, as experimental work has consistently shown that higher loadings increase PdN efficiency and nitrogen removal rates.