Full-Scale N2O Modelling For Advanced Mitigation

Abstract: Effluent quality, performance and nitrous oxide (N2O) emissions of the water resource recovery facility of Hessenpoort (the Netherlands) were investigated using an advanced model-based methodology. Mechanistic activated sludge modelling with N2O kinetics was integrated with Computational Fluid Dynamics (CFD) to design a calibrated and validated CFD-informed dynamic model. The model unraveled the root causes of N2O production and allowed to explore the long-term implications of critical plant changes.

INTRODUCTION
Design and operation of water resource recovery facilities (WRRFs) is becoming a balancing act AMONG effluent requirements, plant efficiency, and minimizing nitrous oxide (N2O) emissions (the '3Es').
N2O emissions from WRRFs can be minimized implementing advanced operational strategies that promote efficient nitrogen removal while minimizing N2O production. The complexity of N2O formation in bioreactors makes it difficult to use general guidelines for measurements and mitigation. While direct onsite monitoring provides estimates of emissions and their dynamics, major limitations are a lack of root cause understanding and limited spatial insights. Integrating sensors, Computational Fluid Dynamics (CFD), and biokinetic modeling enables the development of tailored and effective mitigation strategies along with industry-wide learning.
The Dutch water authority Waterschap Drents Overijsselse Delta (WDODelta) seeked to understand and mitigate N2O emissions at the Hessenpoort WRRF. As part of a larger hydrogen project, in which pure oxygen (PO) is being generated, WDODelta was also investigating the use of pure oxygen aeration given its potential in sensibly reducing N2O stripping, but with rather unknown implications on the biomass (e.g. lowering of pH).
Mechanistic models are well established in the biological water treatment sector and can be used to accelerate decisions on bioreactors' operation as well as N2O sensor placement and N2O mitigation itself.
In this work, a mechanistic activated sludge model integrated with CFD and used to design a CFD-informed dynamic model able to run long dynamic simulations with very detailed representation of the bioreactor.

MATERIALS AND METHODS The biological activity was modelled via the ASM-N2O developed by AM-Team from the available literature (inter alia: Bellandi, 2018, Guo, 2014; Hiatt and Grady, 2008; Peng et al., 2015; Sperandio et al., 2016) (Figure 1).
The hydraulic model used for the CFD simulations was a 2-phase Eulerian model (liquid-air) in which the ASM-N2O was integrated (Figure 2).
The bioreactor consisted of 5 concentric rings (R1-5; Figure 3) operated in a carrousel configuration and interconnected by overflows and 3 recirculatios. The return activated sludge (RAS) and influent entered in R1.

The influent dynamics were monitored with high frequency (1 min) sensors and an influent fractionation of COD (total and soluble), BOD, N species, and P species was performed hourly for 48h to accurately observe influent dynamics. N2O concentrations in the liquid phase (Unisense, Denmark) were applied in R5 (aerobic) and R3 (anoxic). Sensors were regularly calibrated and maintained to assure good data quality.

RESULTS
The CFD integration with the biokinetic ASM-N2O model allowed to run several scenarios among which the use of PO in R4 and R5, which showed a potential reduction of 75 % in N2O emissions. The creation of a CFD-informed tool allowed to run very detailed simulations resembling very closely the N species in the tank (Figure 4). Spatial representation of the bioreactor is a very essential step for being able to accurately predict N2O measurements.
The CFD-biokinetic integration was essential to visualize and understand the gradients in N2O formation, consumption, and transport at different operating conditions (Figure 5, left). Local N2O hotspots in different zones (indicated in red) were the root causes of this WRRFs N2O emissions. Each hotspot was caused by different conditions (e.g. local DO and ammonia gradients). In zones where complete denitrification could take place, N2O consumption could happen (indicated in blue). As a result of the spatial provided to the dynamic model, both in terms of hydrodynamics and influent characterization, it was possible to predict the measured N2O dynamics with the CFD-informed dynamic model (Figure 5, right).

CONCLUSIONS
To assess the current plant performance an initial plant screening was performed using an advanced 3D ASM-N2O model. The reactor's hydrodynamic performance, effluent quality, N2O emissions and plant efficiency were assessed.
The use of PO to limit N2O stripping was simulated indicating the potential of pure oxygen in carbon footprint minimization. Onsite PO tests are currently ongoing.
#The 3D CFD results were used to develop a highly performing dynamic model to be used for future 'what-if' scenario testing. The CFD-informed dynamic model was able to capture carbon and nitrogen dynamics and N2O emissions and is now ready for plant maximisation/modification. Adequate influent data and spatial information were deemed necessary for satisfying N2O predictions.
The dynamic model is now used to test numerous operational and design scenarios for holistic plant optimisation (3E: effluent, efficiency and emissions) and guidance of integration and control in the use of pure oxygen.