Plenary Sessions – Technical Abstracts

Modelling of aerobic granular sludge sequencing batch reactors – Introduction

to the Eawag AGS model
M. Layer1, M. Stähle1, M. Garcia Villodres1, F. Házi2, D. Bencsik1, I. Takács2, E.
Morgenroth1, N. Derlon1
1 Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland ([email protected])
2 Dynamita, 2015 Route d'Aiglun, 06910 Sigale, France ([email protected])

Keywords: Aerobic granular sludge; flocs; mathematical modelling; nutrient removal; plug-flow; settling

Introduction
Aerobic granular sludge (AGS) systems represent a relevant alternative to conventional activated
sludge systems, and a range of AGS models were developed to better understand those systems.
However, the underlying structure of those AGS models was mainly based on results gained from
simple lab-scale systems, resulting in most AGS models not accounting for features relevant in
design and operation of full-scale AGS systems. The simplifications of the existing AGS models
include the following: (1) sludge bed stratification over the reactor height is neglected as the
feeding phase is modelled as completely mixed - variable volume, while full-scale AGS systems
are operated as a plug-flow (PF) - constant volume, (2) flocs are ignored while present in non-
negligible fractions in AGS (10-30%) and thus influencing the effluent quality, (3) the excess
sludge removal and SRT control is not performed as a function of the settling properties and thus
not selective.
As a result, key features of full-scale AGS systems cannot be predicted properly, e.g., the effluent
quality in terms of solids concentration (van Dijk et al., 2018), the denitrification and total nitrogen
removal (Layer et al., 2020b). In order to provide an AGS model relevant for engineers to design,
identify failures or optimize AGS systems, we developed a new model including the following key
features: PF feeding, fully-mixed reaction phases, coexistence of granules and flocs, individual
settling models for flocs and granules and selective excess sludge removal. The current study
provides an overview of those features and demonstrates their importance for the prediction of the
AGS system performances. A specific emphasis is put on demonstrating the relevance of
modelling stratification of the reactor’s heigh during the plug-flow feeding, in comparison to the
Simplified Granular SBR model within SUMO® (no PF, variable volume, no settling, effluent
solids are user input). The different modelling scenarios include (1) the effect of the sludge bed
stratification during PF feeding, (2) effect of upflow feeding velocity (vww) and (3) optimization
of the SBR cycle, e.g., anaerobic mixing / idle phases.

Model implementation
The Eawag AGS model is implemented in SUMO® software and consists of a 1-D biofilm model
(Wanner and Reichert, 1996), a reactor model and a biokinetic model (Sumo1). The reactor model
consists of an assembly of 4 completely stirred tank reactors organised in series, so that PF
conditions is mimicked during feeding at the reactor bottom. The volume of each sub-reactors is
set to perfectly represent bed stratification during PF feeding: granules at the bottom, flocs above
the granules and two smaller sub-reactors above filled with supernatant only. A rule-based
programming based on the specific volume-exchange ratio of each sub-reactor is applied to
improve the PF behaviour. Individual settling models for flocs (flocculated, hindered and

7th IWA Water Resource Recovery Modelling Seminar Page 76 of 225

 Plenary Sessions – Technical Abstracts
compressed settling (Takács et al., 1991) and granules (discrete particle settling, (MWH, 2012)) are
implemented. Excess sludge is withdrawn to specifically achieve a target system SRT or based on
a specific volume removal at a certain time / specific height (resembling wastage of certain full-
scale applications). Only flocs are removed. Effluent XTSS is predicted based on the settling
velocity of flocs (XTSS,bulk) and applied WW upflow velocity (VWW), i.e., if vsett,bulk ≤ vww and if
XTSS,bulk > XTSS,nonsettleable. vww is calculated based on the reactor geometry/inflow and corrected for
the available pore space in the granules bed (bottom sub-reactor).
Spherical granules divided in n layers are considered in the 1-D biofilm model. The thickness of
the 4 outer layers is fixed to 25 µm to increase the resolution of the mass-transfer processes at the
granule surface, and allow accurate predictions of the total nitrogen removal (Layer et al., 2020a).
Only one granule diameter (input variable) and a constant granules number are considered. The
density of each granule layer is a function of mass-transfer and growth mechanisms. Mass-transfer
mechanisms include (1) diffusion of soluble/colloidal compounds, (2) displacement of particulate
compounds, (3) attachment of particulate compounds, and (4) internal transfer of particulate
compounds within the granule compartments.
The Sumo1 biokinetic model used in this study includes the following processes: (1) COD
removal, (2) 1-step nitrification-denitrification, (3) fermentation and (4) enhanced biological
phosphorus removal (EBPR) based on PAO - GAO competition.
The default scenario considered default biofilm and reactor parameters. The granules were
characterised by n=6 layers and a granule diameter of 1.5 mm. The SBR cycle was characterised
by 1.5 h anaerobic PF feeding (vww = 1.7 m h-1, volume exchange ratio (VER) = 0.5, standard
SUMO® municipal WW composition), 4.5 h aerobic fully-mixed conditions (dissolved oxygen
concentration in the bulk = 2.0 mgO2 L-1), 0.5 h settling and wasting (target SRT = 20 d). A set of
scenarios was performed to illustrate the functionalities of the Eawag AGS model: (1) vww
scenarios (1, 1.7, 2.5, 5 m h-1) at constant VER; the shortened anaerobic PF phase is compensated
for by a longer aerobic SBR phase and (2) optimisation scenarios (anaerobic PF feeding (60,
120 min) + anaerobic mixing (AM) (60, 0 min) or anaerobic idle (ID) (60, 0 min) phases). Data of
one SBR cycle was extracted after 100 d of simulation and analysed using R (Version 3.6.3, 2020).

Results and Discussion

Predicting stratification over the reactor height is key. Our model allows predicting a
characteristic distribution of granules / flocs over the reactor height during settling and plug-flow
feeding (Figure 1A, 1B). The stratification of the sludge bed during feeding results from the
specific settling properties of the granules / flocs and applied vWW. Granules quickly settle towards
the bottom of the reactor while flocs accumulate in the bottom-up sub-reactor during settling. No
flocs other than the non-settleable XTSS fraction (20 mgTSS L-1) are present in the top sub-reactor
after settling and thus washed out of the SBR once plug-flow feeding is applied. Such stratification
of the sludge bed during settling / feeding is representative of our pilot / full-scale observations
(Nereda® pilot tests). The Eawag AGS model also allows predicting the distribution of soluble
and particulate substrates / nutrient over the reactor height (Figure 1, D and E). We could
demonstrate that predicting the stratification of the sludge bed has a key influence on the substrate
/ nutrient microbial conversion processes (Figure 1D, 1E, 1F). Our model predicts a highly
competitive advantage of the biomass located at the bottom of the reactor and in direct contact
with fresh WW, i.e., the granules. Modelling the stratification of the sludge bed allows predicting
fast denitrification and rapid establishment of anaerobic conditions, resulting in a larger

7th IWA Water Resource Recovery Modelling Seminar Page 77 of 225

 Plenary Sessions – Technical Abstracts
accumulation of PAO than when the AGS system is fully mixed (highlighted by a much larger
release of PO43-).

Figure 1 Distribution of (A) granules, (B) flocs, (D) nitrates and (E) ortho-phosphates predicted by the Eawag AGS
model. Distribution of (C) flocs and granules and (F) nitrate and ortho-phosphate within the Simplified Granular SBR
model.
The Eawag AGS model is a relevant tool to better understand AGS systems. One relevant
example is the effect of the influent WW upflow velocity vww that directly governs the effluent
quality. Low vww value of 1 to 4 m h-1 are optimal for the selection of carbon-storing organisms
(PAO and GAO), and for the effluent quality in terms of total nitrogen and solids (XTSS). In this
range, the lower the value of vww the lower the effluent nitrate concentrations and the larger the
storage of poly-phosphate. Increasing vww results in an increasing wash-out of XTSS with the
effluent. At 4 to 6 m h-1 (vww≈vsett,flocs) the effluent XTSS decreases from 20-30 mg L-1 at the
beginning of the feeding to the non-settleable XTSS concentration (20 mg L-1) at the end. For vww
> 6 m h-1 (vww>vsett,flocs) large effluent XTSS concentrations of >1000 mg L-1 are predicted.
The Eawag AGS model is a relevant tool for optimising the SBR operation of AGS system.
Understanding how to optimize SBR cycles in order to improve the system performances require
a deep understanding of the system and is experimentally heavy. A relevant approach is to first
use a model. An example of how to use the Eawag AGS model to optimise AGS systems is the
introduction of AM or ID phases (1 h) after a shortened PF feeding phase (1 h). AM and ID phases
help to increase the total nitrogen removal and led to increased poly-phosphate release and
anaerobic utilisation of organic substrates in comparison to the default scenario (1.5 h PF) or
prolonged PF feeding (2 h). The Eawag AGS model thus represents a valuable tool for process
optimisation.

7th IWA Water Resource Recovery Modelling Seminar Page 78 of 225

 Plenary Sessions – Technical Abstracts
Conclusion
• Stratification over the height of the reactor during filling matters! Modelling plug-flow feeding
is key to predict the behaviour and performances of AGS-SBR. The stratification of the sludge
bed and of the growth conditions governs the microbial community composition, and
ultimately the microbial transformation and effluent quality.
• The Eawag AGS model is a relevant tool to better understand AGS systems. AGS systems
applied for the treatment of municipal WW are often considered as a black-box. The current
model can be used as a mechanistic tool to better understand those systems, e.g., the
development of gradient of growth conditions within the sludge bed and its effect on the
effluent quality, the role of flocs in the substrate / nutrient turn-over.
• The Eawag AGS model is a relevant tool for the optimisation and design of AGS-SBR. Many
aspects of the SBR operation can be tweaked in order to optimize the system performances,
e.g., the anaerobic phase (mixed or idle), the feeding upflow velocity, the aeration strategy.
Evaluating the effect of those different variables is experimentally difficult. Our model can be
used as a tool to browse through different scenarios and identify promising routes of
optimisation, before full-scale evaluation.

Our opinion regarding the modelling of AGS-SBR systems

A main feature of biofilm systems is the development of concentration gradients over the biofilm
thickness, that in turn defines the local growth conditions and overall system performances. But
for AGS-SBR operated at constant volume, a stratification of the growth conditions also develops
over the reactor height, as a result of the PF feeding and bed stratification. We advocate that an
accurate prediction of the performances of AGS-SBR require a rigorous prediction of the
concentration gradients over both the reactor height and granules radius. These aspects are
especially relevant for predicting the microbial conversion rates and also the effluent quality. We
argue that models that are not predicting the concentration gradients over the reactor height are not
suitable tools to predict the performance of AGS-SBR systems operated at constant volume.

References
Layer, M., Garcia Villodres, M., Hernandez, A., Reynaert, E., Morgenroth, E. and Derlon, N. 2020a.
Limited simulatenous nitrification-denitrification (SND) in aerobic granular sludge systems treating
municipal wastewater: Mechanisms and practical implications. Water Research X (7), 100048.
Layer, M., Villodres, M.G., Hernandez, A., Reynaert, E., Morgenroth, E. and Derlon, N. 2020b. Limited
simultaneous nitrification-denitrification (SND) in aerobic granular sludge systems treating municipal
wastewater: Mechanisms and practical implications. Water Research X 7.
MWH (2012) Water treatment: principles and design, John Wiley & Sons.
Takács, I., Patry, G.G. and Nolasco, D. 1991. A dynamic model of the clarification-thickening process.
Water research 25(10), 1263-1271.
van Dijk, E.J.H., Pronk, M. and van Loosdrecht, M.C.M. 2018. Controlling effluent suspended solids in the
aerobic granular sludge process. Water Research 147, 50-59.
Wanner, O. and Reichert, P. 1996. Mathematical modeling of mixed-culture biofilms. Biotechnol Bioeng
49(2), 172-184.

7th IWA Water Resource Recovery Modelling Seminar Page 79 of 225