WEFTEC 2010

Verification of WWTP design guidelines with activated sludge

process models

Lluís Corominas1.2, Xavier Flores-Alsina1, Dirk Muschalla1, Marc B. Neumann1 and

Peter A. Vanrolleghem1*
1
modelEAU, Département de génie civil et génie des eaux, Université Laval, 1065 av. de la
Médecine, Québec (QC) G1V 0A6, Canada
2
ICRA, Catalan Institute for Water Research, Emili Grahit 101, E- 17003 Girona (Spain)

* Corresponding author: peter.vanrolleghem@gci.ulaval.ca.

ABSTRACT
There is significant interest in using process models for WWTP design to complement the
traditional design guidelines. However, transparent quantification of the differences in the design
results obtained with the two approaches has not been done yet. The objective of this paper is to
define a methodology to evaluate the expected performance of a designed WWTP. The use of
this methodology will enable to identify inherent safety incorporated in design guidelines by
verification with dynamic process models. The methodology includes two parts: first, multiple
designs are made using a design guideline for different combinations of influent characteristics,
effluent requirements, safety factors and operating preferences. Second, the results of the designs
(e.g. reactor volumes) are applied in a process model that emulates WWTP behaviour. The
effluent concentrations predicted by the process model are then compared to the effluent
requirements which were imposed for the designs. The methodology is illustrated using the
Metcalf & Eddy design guidelines for a nitrogen-removing process and the ASM1 process
model.

KEYWORDS: Activated sludge models Design, Monte Carlo, Nitrogen removal, Uncertainty,
Wastewater treatment

INTRODUCTION

The use of activated sludge models (ASM) (Henze et al., 2000) is common practice within the
field of wastewater treatment (Copp et al., 2009; Phillips et al., 2009). These models have been
widely used for learning, design and process optimisation purposes (Gernaey et al., 2004).
Focusing on the design applications these ASM-type models have been helpful to evaluate
different alternatives for designing and upgrading wastewater treatment plants (WWTP)
(Daigger and Nolasco, 1995; Hao et al., 2001; Salem et al., 2002; Larrea et al., 2007), to
optimize the design exercises (Rivas et al., 2008), to evaluate costs (Vanrolleghem et al., 1996),
to extract knowledge (Flores et al., 2006), to evaluate the production of Green-house gases
(Johnson and Hiatt, 2009) and to study uncertainty related to the process design (Sin et al.,
2009).

Copyright ©2010 Water Environment Federation. All Rights Reserved.

137
 WEFTEC 2010

Guidelines traditionally used for design (Metcalf & Eddy, ATV, Grady, Ten State Standards,
HSA principles) are based on simplified system descriptions. In recent years there is a growing
interest in applying the dynamic ASM-type (chemical-biological) process models to complement
the traditional design guidelines. For instance, the HSG group in Germany presented a study
(Spering et al., 2008) where the ASM3 model parameters were adjusted to obtain similar design
results compared to the ATV design guidelines. However, there is no clear methodology to
evaluate the performance of traditional design guidelines using the ASM-type models.

The objective of this paper is to develop a methodology that is able to estimate the reserve
capacity inherent in a design guideline. The methodology is based upon comparing the results of
a certain guideline to a reference. In this case it is proposed to use the ASM1 model as reference.
The methodology developed in this paper should be applicable to other design guidelines.

METHODOLOGY

The methodology suggested for assessing the reserve capacity inherent in a design guideline is
presented in Figure 1 and explained below.

Safety factors
Design
Influent
Design outcomes: 
characteristics
Design ‐ Contruction volumes
guideline ‐ Air blower capacity
Operating ‐ Pumping capacity
preferences

Effluent
requirements

? Evaluation
Effluent
concentrations

Process model

Figure 1. Methodology for design guideline evaluation

Multiple designs generation using guidelines

Designs for a given guideline are determined by defining influent wastewater characteristics,
specifying operating preferences (e.g. DO and MLSS concentration in the reactors), selecting
safety factors (SF) and setting the effluent requirements. The design outcomes are the reactor
volumes, air blower capacity and pumping capacity. A full exploration of the «design space» of a

Copyright ©2010 Water Environment Federation. All Rights Reserved.

138
 WEFTEC 2010

guideline is conducted by repeating this exercise for many possible combinations of these design
conditions. This is achieved by selecting a range of values for the design conditions, in this study
through uniform probability density functions. The Monte Carlo (MC) simulation technique
using efficient Latin Hypercube sampling (LHS) is applied to propagate the ranges of the design
conditions to the design outcomes.

Evaluation of the design guideline using dynamic process models

The lower part of Figure 1 shows the evaluation step. The same inputs used for the designs
(influent characteristics and operating preferences, e.g. DO and MLSS concentration in the
reactors) together with the design outcomes obtained from the guideline are used as inputs to the
process model. The predicted effluent concentrations obtained with the process model are then
compared to the effluent requirements which were imposed for the designs. This enables to
investigate whether the guidelines lead to optimal designs or to over- or under-sized plants. It is
important to stress that the same criteria used for the design requirements (e.g. weekly average of
effluent concentrations) are applied in the evaluation step.
Case study
In this case study, the Metcalf & Eddy guidelines (Tchobanoglous et al., 2003) are used to size a
modified Ludzack-Ettinger plant for nitrogen removal. In order to explore the «design space»,
1000 design configurations are created by random sampling from uniform probability density
functions for the influent fractions, the effluent requirements, the safety factors and the dissolved
oxygen operating concentration. The uniform probability density functions are characterised by
the lower and upper values in Table 1.
Table 1. Range of values of design conditions expressed as uniform distributions characterised by default,
upper and lower values

Design conditions Symbol Default Lower value Upper value Units

Influent fractions
Undegradable soluble f_SU 0.09 0.05 0.14 -
Biodegradable soluble f_SB 0.16 0.08 0.24 -
Undegradable particulates f_XU,inf 0.12 0.06 0.18 -
Biodegradable particulates f_XB 0.52 0.35 0.72 -
Heterotrophic biomass f_XOHO 0.11 0.06 0.17 -
Effluent requirements (weekly average)
Effluent ammonium SNHx 2 0.5 6 gN·m-3
Effluent nitrate SNOx 6 5 10 gN·m-3
Safety factors
Aerobic section SFAER 1.25 1 1.5 -
Anoxic section SFANOX 1.25 1 1.5 -
Operational conditions
Dissolved oxygen (aerobic zone) SO2 2 0.5 4 (-gCOD)·m-3

The time series of influent characteristics and the parameter values for the dynamic simulations
are taken from Copp (2002) which considers an average influent dry-weather flow rate of 18446
m3·d-1, an average biodegradable COD of 300 g COD·m-3 and an average ammonium

Copyright ©2010 Water Environment Federation. All Rights Reserved.

139
 WEFTEC 2010

concentration of 50 g NH4-N·m-3. Taking this influent data as starting point 1000 influent files
are generated applying different coefficients for the influent fractionation of the organic matter.
The total COD load profile is the same for the 1000 influent files, but the fractions of inorganic
soluble (f_SU), organic biodegradable (f_SB), undegradable particulates (f_XU,inf) and the
heterotrophic biomass (f_XOHO) fractions are randomly sampled according to Table 1. For each
of the files the mass balance is closed by assigning the rest of COD to the fraction of the organic
particulates XB (f_XB = 1-f_SU – f_SB – f_XU,inf – f_XOHO).
Effluent requirements are defined as average weekly concentrations and the ranges are assumed
to cover values that would typically be envisaged by regulators.
The process model used to evaluate the designs includes the biokinetic Activated Sludge Model
nº1 (ASM1, Henze et al., 2000) and the settler model defined in Takács et al. (1991). A
dissolved oxygen controller in the aerobic reactor and a MLSS controller for all reactors have
been implemented in order to impose the operating preferences of the design on the dynamic
process model simulations. The simulation procedure is based on 150 days of steady-state
simulation, followed by 14 days of dynamic simulation and another final 14 days of dynamic
simulation. The evaluation of the simulation results is based on the average of the last 7 days of
dynamic simulation. This simulation procedure is conducted for each of the designs obtained
with the guideline (the overall balance is 1000 simulations).

RESULTS AND DISCUSSION

Designs obtained with Metcalf & Eddy guideline

The results presented in this paper focus on the design of the aerobic reactor. The histogram of
the 1000 designed volumes presented in Figure 2 is skewed with values ranging between 4000
m3 and 17000 m3. This range of values is quite large, demonstrating that a wide range of design
possibilities has been explored.

250

200

150
Count

100

50

0
4000 6000 8000 10000 12000 14000 16000 18000
3
Aerobic volume (m )

Figure 2. Histograms of the aerobic volumes for the 1000 designs using the Metcalf & Eddy guideline

The variation of the aerobic volume with respect to design conditions (DO set-point and effluent
ammonia) (see Table 1) is presented in Figure 3. It can be observed that the design volume is

Copyright ©2010 Water Environment Federation. All Rights Reserved.

140
 WEFTEC 2010

sensitive to the desired effluent ammonia. This was validated with a sensitivity analysis
presented in Flores-Alsina et al. (2010) where the effluent ammonia concentration was found to
be the most influential input with respect to the aerobic volume. The influence of the safety
factors on the volume can also be seen in Figure 3. The higher the safety factors the lower the
volumes obtained.
12000
12000 SF [1.00-1.16]
10000 SF [1.16-1.33]

V AEROBIC (m )
SF [1.33-1.50]

3
10000
8000
V AEROBIC (m )
3

8000 6000

4000

)
t (gN.m -3
6000

DO
2000
0

in
1 4
4000 2 3

AER
SF [1.00-1.16] 3

4 effluen
SF [1.16-1.33] 4 2
5 1

(g.m
SF [1.33-1.50] 0
2000 6 3 2 1
4 7 0 5 4
6

-3 )
3 2 7
NH +

1 0
-3
+ nt (gN.m )
DO in AER (g.m-3) NH4 efflue

Figure 3. Results of the sizing of the aerobic reactor for the 1000 designs using the Metcalf & Eddy
guideline

Evaluation of designs with a dynamic process model

The goal of this study is the evaluation of a design guideline with an ASM1-based process
model. In order to evaluate the design results for the aerobic volume, the effluent NH4+
requirements for each design are plotted against the predicted concentration (effluent NH4+)
obtained from the process model, as presented in Figure 4.
6
SF [1.00-1.16]
SF [1.16-1.33]
5 SF [1.33-1.50]
Predicted NH4+ (gN.m-3)

0
0 1 2 3 4 5 6
Design NH4+ (gN.m-3)

Figure 4. Predicted concentrations obtained with the ASM1 process model vs. the Metcalf & Eddy design
effluent requirement for the 1000 designs.

Copyright ©2010 Water Environment Federation. All Rights Reserved.

141
 WEFTEC 2010

It can be observed that there is a monotonous increasing relationship between the imposed
effluent requirements for the Metcalf & Eddy designs and the concentrations predicted with the
process model. Figure 4 indicates that the aerobic reactor volumes, obtained with the Metcalf &
Eddy guideline, are oversized i.e. the predicted ammonia concentration values (from the dynamic
simulations) are (significantly) below the concentration requirements imposed for the design.
Even the use of low safety factors leads to overdesign.

Discussion of differences observed

In order to understand the differences obtained for the design and predicted effluent ammonia
several factors have to be taken into account.

1) Predicted SRTs with the BSM1 model are higher than the chosen SRTs during the designs
(see Figure 5). In fact, the design allows selecting two important characteristics of the treatment
plant: the MLSS concentration in the bioreactor and the SRT. However, it is impossible to
impose both on the dynamic process model. Both options were tested and it turned out that
imposing the design MLSS concentration on the dynamic simulations led to a less extreme
sludge production (see Figure 6). By imposing an MLSS concentration, the BSM1 simulations
result in higher SRTs and therefore higher nitrification capacities are obtained, explaining the
lower ammonium concentration in the effluent. The differences in the SRTs can be explained by
the different model structures of both models (see equation 1 and equation 2) that do not allow
fixing the MLSS and the SRT of the system at the same time. Note that fixing the SRTs (and
leaving the MLSS to evolve to lower values) would make that the effluent ammonia
concentrations would come closer to the values the plant was designed for, but the sludge
production results become quite extreme (results not shown).
30

25
Predicted SRT (d)

20

15

10

0
0 5 10 15 20 25 30
Design SRT (d)
Figure 5. Design vs simulated SRT for the 1000 design exercises

SRT calculation Metcalf & Eddy (Eq 1) SRT calculation BSM1 (Eq 2)
1

,
NHx,ANO O2,ANO

Copyright ©2010 Water Environment Federation. All Rights Reserved.

142
 WEFTEC 2010

2) Sludge production is much higher in the BSM1 model simulations compared to the expected
sludge production from the Metcalf & Eddy design (Figure 6).

Predicted sludge production (kgVSS.d-1)

4000

3500

3000

2500

2000

1500

1000
1000 1500 2000 2500 3000 3500 4000
Design sludge production (kgVSS.d-1)

Figure 6. Design vs simulated sludge production for the 1000 design exercises

This difference can be explained by the more detailed BSM1 model structure, coupled to
different default kinetic and stoichiometric parameters for both Metcalf & Eddy guidelines and
BSM1 (see Table 2). An example of this is the death-regeneration concept that is considered in
the ASM1 models, that leads to different meaning for the decay rates and the undegradable
particulates (XU) generated in biomass decay. However, for these parameters it is well-known
how to make decay rates and the fraction of undegradable particulates comparable for the two
model structures (see equation 3 and equation 4 where parameters with a prime are the Metcalf
& Eddy parameters).

Table 2. Parameter values used in the Metcalf & Eddy design (values at 15ºC). Notation taken from
Corominas et al. (2010)

Parameter Symbol Metcalf & BSM1

Eddy
Maximum growth rate of XOHO μOHO,Max 4.27 4
Half-saturation coefficient for SB KSB,OHO 20 10
Decay rate for XOHO b’OHO 0.098 0.115 (bOHO=0.3)
Maximum growth rate of XANO μANO,Max 0.3813 0.5
Half-saturation coefficient for SO2 KO2,ANO 0.5 0.2
Half-saturation coefficient for SNHx KNHx,ANO 0.4415 1
Decay rate for XANO b’ANO 0.0658 0.019 (bANO=0.05)
Yield for XOHO growth YOHO 0.4 0.67
Yield of XANO growth per SNOx YANO 0.12 0.24
Fraction of XU generated in biomass decay f’XU_Bio,lys 0.15 0.20 (fXU_Bio,lys=0.08)

XU_Bio,lys
Eq 3 XU_Bio,lys
Eq 4
1 1 XU_Bio,lys
XU_Bio,lys

Copyright ©2010 Water Environment Federation. All Rights Reserved.

143
 WEFTEC 2010

On the other hand, for other parameters it is not that simple. For instance, the observed yields in
Metcalf & Eddy cannot be compared to the yield used in the BSM1 model.

The sensitivity of the Metcalf & Eddy volumes to the kinetic and stoichiometric parameters will
be explored in future work by following the methodology described in Flores et al. (2010).

Evaluation of the design guideline verification methodology

The proposed methodology can be used to quantify the reserve capacity inherent in a design
guideline, e.g. by how much can the reactor volume be reduced until the predicted effluent
concentrations reach the design effluent requirements. Such calculation is conducted by selecting
several designs and for each of them running different dynamic simulations applying gradual
changes in the volume of the aerobic reactors. An example for one of the 1000 designs is
presented in Figure 7. For this specific case, the ASM-type model shows that the aerobic volume
could be reduced by 35% compared to the design of Metcalf & Eddy. When removing the safety
margins included in the dimensioning guidelines the plant size can be reduced significantly. In
Benedetti et al. (2010) the volumes obtained with ATV design guidelines were reduced up to
60% of its original volume when using a dynamic model.

30

25
Predicted effluent NH4+ (mg·L-1)

20

15

10

Design effluent NH4+

5

0
0 5 10 15 20 25 30 35 40 45 50 55

Volume reduction (%)

Figure 7. Predicted effluent ammonia at different volume reductions for one design exercise (SF=1.33;
DO set-point = 2.87 mgO2·L-1; effluent ammonia = 5.5 mgN·L-1; Aerobic volume = 4830 m3)

The guideline verification methodology proposed here can be applied to different influent
characteristics, operating preferences, safety factors, effluent requirements and kinetic and
stoichiometric parameters in order to explore a full range of design options. This is conducted by
performing a large number of design verifications (1000 Metcalf & Eddy designs and 1000
dynamic simulations in this illustrative case). This methodology is also complemented with a
sensitivity analysis of the results obtained with the design guideline.

Copyright ©2010 Water Environment Federation. All Rights Reserved.

144
 WEFTEC 2010

With this study the authors also want to make the point that modifying the ASM-type models to
describe the results obtained by a specific guideline as performed by Spering et al. (2008) will
not improve the design guidelines themselves. Both design guidelines and dynamic process
model simulations have their role to play. One should only be aware of the differences and get
some idea on the inherent reserve capacity present.

This study demonstrates that the verification of design guidelines is a complex issue and there is
still further work to do. Ongoing research focuses on studying the underlying mechanisms that
are behind the different model structures and to find the best approaches for objective
comparison. The proposed methodology will also be applied to other design guidelines for cross-
comparison of inherent safety.

CONCLUSIONS

A methodology is proposed and validated to verify the results obtained from different design
guidelines by using ASM-type model simulations. The methodology suggests i) to perform
multiple design exercises for a given guideline applying different ranges for design conditions
and influent wastewater fractionation, ii) to run multiple ASM-type dynamic model simulations
using the outcomes of the design exercises as inputs and iii) to compare the resulting effluent
ammonia, sludge production, etc. The results show that the Metcalf & Eddy guidelines give
over-sized aerobic reactors compared to the BSM1 simulations due to the differences in the
model structure and their default model parameters. The different structures of the models make
the verification of the design guidelines with a dynamic model a complex problem. For this
reason, the verification has to be conducted by checking different criteria, including, for the
illustrative case, the effluent ammonia, the SRTs and the sludge production obtained with the
different approaches.

ACKNOWLEDGMENTS

Peter Vanrolleghem holds the Canada Research Chair on Water Quality Modeling. Lluís
Corominas received the scholarship “Juan de la Cierva” from the Spanish Ministry. The authors
wish to acknowledge the support provided by the IWA Task Group on Design and Operations
Uncertainty.

REFERENCES
Benedetti L., De Keyser W., Nopens I. and Vanrolleghem P.A. (2010) Probabilistic modelling
and evaluation of wastewater treatment plant upgrades in a water quality based evaluation
context. Journal of Hydroinformatics, 12(4), 380-395.
Copp J.B. (2002) The COST Simulation Benchmark: Description and Simulator Manual. Office
for Official Publications of the European Community, Luxembourg.
Copp J.B., Johnson B.R., Shaw A., Burbano M.S., Narayanan B., Frank K., Kinnear D., Melcer
H. and Brischke K. (2009) A balancing act: the consulting engineers' pragmatic view of
process modelling. Water Sci. Technol., 59(4), 763-769.
Daigger G.T. and Nolasco D. (1995) Evaluation and design of full-scale wastewater treatment
plants using biological process models. Wat. Sci. Technol., 31(2), 245-255.

Copyright ©2010 Water Environment Federation. All Rights Reserved.

145
 WEFTEC 2010

Flores X., Poch M., Rodríguez-Roda I., Jiménez L. and Bañares-Alcántara R. (2006) Knowledge
extraction during the design of activated sludge systems. Computer Aided Chemical
Engineering 21, 1083-1088.
Flores-Alsina, X. Corominas Ll., Muschalla D., Neumann M.B. and Vanrolleghem P.A. (2010)
How do Initial Design Assumptions Determine Plant Sizing? Assessing Activated Sludge
Process Design using Monte-Carlo Simulation and Global Sensitivity Analysis. In
proceedings: IWA World Water Congress and Exhibition, 19–24 September 2010,
Montréal, Canada.
Gernaey K.V., Van Loosdrecht M.C.M., Henze M., Lind M. and Jorgensen S.B. (2004)
Activated sludge wastewater treatment plant modelling and simulation: state of the art.
Environmental Modelling & Software, 19, 763-783.
Hao X., van Loosdrecht M.C.M., Meijer S.C.F. and Qian Y. (2001) Model-based evaluation of
two BNR processes-UCT and A2N. Water Res., 35, 2851–2860.
Henze M., Gujer W., Mino T. and van Loosdrecht M.C.M. (2000) Activated Sludge Models:
ASM1, ASM2, ASM2d and ASM3. Scientific and Technical Report nº9. IWA Publishing,
London, UK.
Johnson B.R. and Hiatt W.C. (2009) Nitrogen removal system impacts on secondary treatment
greenhouse gas production and whole plant carbon footprint. In proceedings: Nutrient
removal 2009, June 28 - July 1, Washington, DC (US).
Larrea L., Albizuri J., Irizar I. and Hernández J.M. (2007) Design and operation of SBR
processes for small plants based on simulations. Water Sci. Technol., 55(7), 163-171.
Phillips H.M., Sahlstedt K.E., Frank K., Bratby J., Brennan W., Rogowski S., Pier D., Anderson
W., Mulas M., Copp J.B. and Shirodkar N. (2009) Wastewater treatment modelling in
practice: a collaborative discussion of the state of the art. Water Sci. Technol., 59(4),
695–704.
Rivas A., Irizar I., Ayesa E. (2008) Model-based optimisation of wastewater treatment plants
design. Environmental Modelling & Software, 23, 435-450.
Salem S., Berends D., Heijnen J.J., van Loosdrecht M.C.M. (2002) Model-based evaluation of a
new upgrading concept for N-removal. Water Sci. Technol., 45 (6), 169–176.
Sin G., Gernaey K.V., Neumann M.B., van Loosdrecht M.C.M. and Gujer W. (2009)
Uncertainty analysis in WWTP model applications: A critical discussion using an
example from design. Water Res., 43(11), 2894-2906.
Spering V., Alex J., Langergraber G., Ahnert M., Halft N., Hobus I., Weissenbacher N., Winkler
S. and Yücesoy E. (2008) Using dynamic simulation for design of activated sludge plants.
In: Proceedings International Symposium on Sanitary and Environmental Engineering
(SIDISA.08), 24-27 June 2008, Florence, Italy. (CD-ROM, paper 307).
Takács I., Patry G.G. and Nolasco D. (1991) A dynamic model of the clarification-thickening
process. Water Res., 25(10), 1263.
Tchobanoglous G, Burton FL and Stensel HD (2003) Wastewater Enginnering: Treatment,
Disposal and Reuse. McGraw-Hill Higher Education, New York.
Vanrolleghem P.A., Jeppsson U., Carstensen J., Carlsson B. and Olsson G. (1996) Integration of
wastewater treatment plant design and operation – a systematic approach using cost
functions. Water Sci. Technol., 34(3-4), 159-171.

Copyright ©2010 Water Environment Federation. All Rights Reserved.

146