See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/49844360

BSM-MBR: A benchmark simulation model to

compare control and operational strategies for
membrane bioreactors

Article in Water Research · March 2011

DOI: 10.1016/j.watres.2011.01.006 · Source: PubMed

CITATIONS READS

41 131

5 authors, including:

Thomas Maere Simon jon Judd

Laval University Qatar University
64 PUBLICATIONS 342 CITATIONS 189 PUBLICATIONS 6,662 CITATIONS

SEE PROFILE SEE PROFILE

Ingmar Nopens
Ghent University
309 PUBLICATIONS 2,904 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Innovative nutrient regulations for wastewater treatment plants (InnovaReg) View project

Full-scale calibration of the Bürger-Diehl settler modelling framework View project

All content following this page was uploaded by Thomas Maere on 02 August 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

BSM-MBR: A benchmark simulation model to compare control

and operational strategies for membrane bioreactors

Thomas Maere a,*, Bart Verrecht b, Stefanie Moerenhout a, Simon Judd b, Ingmar Nopens a
a
BIOMATH, Department of Applied Mathematics, Biometrics and Process Control, Ghent University,
Coupure Links 653, B-9000 Gent, Belgium
b
School of Water Sciences, Cranfield University, SIMS, Building 52, Cranfield, Bedfordshire MK43 0AL, UK

article info abstract

Article history: A benchmark simulation model for membrane bioreactors (BSM-MBR) was developed to
Received 29 September 2010 evaluate operational and control strategies in terms of effluent quality and operational
Received in revised form costs. The configuration of the existing BSM1 for conventional wastewater treatment
7 January 2011 plants was adapted using reactor volumes, pumped sludge flows and membrane filtration
Accepted 10 January 2011 for the water-sludge separation. The BSM1 performance criteria were extended for an MBR
Available online 18 January 2011 taking into account additional pumping requirements for permeate production and aera-
tion requirements for membrane fouling prevention. To incorporate the effects of elevated
Keywords: sludge concentrations on aeration efficiency and costs a dedicated aeration model was
BSM adopted. Steady-state and dynamic simulations revealed BSM-MBR, as expected, to out-
Control perform BSM1 for effluent quality, mainly due to complete retention of solids and
MBR improved ammonium removal from extensive aeration combined with higher biomass
Modelling levels. However, this was at the expense of significantly higher operational costs. A
Operational cost comparison with three large-scale MBRs showed BSM-MBR energy costs to be realistic. The
Optimization membrane aeration costs for the open loop simulations were rather high, attributed to
non-optimization of BSM-MBR. As proof of concept two closed loop simulations were run
to demonstrate the usefulness of BSM-MBR for identifying control strategies to lower
operational costs without compromising effluent quality.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction operational costs can be hugely influential in determining

selection of MBRs, particularly at large-scale.
The use of membrane bioreactors (MBRs) for wastewater For conventional wastewater treatment plants (WWTPs)
treatment has increased significantly over the last 15 years and MBRs, mathematical models like the ASM family (Henze
due to technological advances and generally decreasing et al., 2000) are widely used for studying process behavior,
membrane costs. The high effluent quality offered compared system design and process optimization (Fenu et al., 2010a;
to conventional activated sludge (CAS) systems makes MBRs Gernaey et al., 2004; Verrecht et al., 2010a). The latter has
especially suited for reuse (Judd, 2008). Their widespread also led to the development of dedicated tools such as the
application, however, is still limited by comparatively high COST/IWA Benchmark Simulation Model No. 1 (BSM1) (Copp,
life cycle costs over more conventional available options 2002; Jeppsson and Pons, 2004), which is a standardised
(Kinnear et al., 2010). Marginal decreases in both capital and simulation procedure for the design and evaluation of control

* Corresponding author. Tel.: þ32 92645937; fax: þ32 92646220.

E-mail addresses: thomas.maere@ugent.be (T. Maere), bart.verrecht@thameswater.co.uk (B. Verrecht), stefanie.moerenhout@ugent.be
(S. Moerenhout), s.j.judd@cranfield.ac.uk (S. Judd), ingmar.nopens@ugent.be (I. Nopens).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.01.006
2182 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

List of symbols and abbreviations PF_Qx pumping energy factor for sludge flow x
1 (kWh m3)
AE aeration energy (kWh d )
PEsludge contribution to pumping energy by all sludge
AEbioreactor contribution to aeration energy by fine bubble
flows (kWh d1)
aeration (kWh d1)
pin absolute inlet pressure (Pa)
AEmembrane contribution to aeration energy by coarse bubble
pout absolute outlet pressure (Pa)
aeration (kWh d1)
PU pollution unit ()
AEtotal total aeration energy (kWh d1)
QA airflow rate (Nm3 d1)
AOTE actual oxygen transfer efficiency (%)
Qe effluent flow rate (m3 d1)
ASM activated sludge model
Qi influent flow rate (m3 d1)
ASM1 activated sludge model no. 1
Qi,av average influent flow rate (m3 d1)
BOD5 5-day biological oxygen demand (g m3)
Qi,max peak instantaneous influent flow rate (m3 d1)
BSM1 benchmark simulation model no. 1
Qint internal nitrate recirculation flow rate (m3 d1)
BSM1_LT long-term benchmark simulation model no. 1
Qr return activated sludge flow rate (m3 d1)
BSM2 benchmark simulation model no. 2
Qw waste flow rate (m3 d1)
BSM-MBR benchmark simulation model for membrane
R universal gas constant (J mol1 K1)
bioreactors
SADm specific membrane aeration demand per unit of
C actual oxygen concentration in the aeration tank
membrane area (Nm3 h1 m2)
(g m3)
SALK alkalinity concentration (molHCO 3 m )
3
CAS conventional activated sludge
SI soluble inert organic material concentration
COD chemical oxygen demand (g m3)
(gCOD m3)
cp power factor (kWs kg1)
SND soluble biodegradable organic nitrogen
C*(20) dissolved oxygen saturation concentration in
concentration (gN m3)
clean water at 20  C and 1 atm (g m3)
SNH ammonia plus ammonium nitrogen
cSI constant for unit conversion ()
concentration (gN m3)
C*(T) dissolved oxygen saturation concentration for
SNH,limit_violations number of exceedances of effluent SNH over
clean water at temperature T at sea level (g m3)
,av 4 gN m3 ()
C* (T) average dissolved oxygen saturation
SNH,95 95th percentile for effluent SNH (gN m3)
concentration for clean water in an aeration tank
SNO nitrite plus nitrate nitrogen concentration (gN m3)
for a given temperature T at sea level (g m3)
SO dissolved oxygen concentration (g m3)
DO dissolved oxygen concentration (g m3)
SOTE standard oxygen transfer efficiency (% m1)
e blower efficiency ()
SP sludge production for disposal (kgTSS d1)
EQI effluent quality index (kgPU d1)
SPtotal total sludge production (kgTSS d1)
F correction factor for fouling of the air diffusers
SRT sludge retention time (d1)
(1 for clean diffusers)
SS soluble, readily biodegradable organic material
g gravitational acceleration (m s2)
concentration (gCOD m3)
h depth of the aeration tank (m)
t time (d)
HRT hydraulic retention time (h1)
T temperature of the mixed liquor ( C)
IQI influent quality index (kgPU d1)
Tev evaluation period (d)
LMH unit for flux, i.e. l m2 h1
Tin absolute inlet temperature (K)
MBR membrane bioreactor
TKN total Kjeldahl nitrogen concentration (g m3)
ME mixing energy (kWh d1)
TN total nitrogen concentration (gN m3)
n air constant ()
TNlimit_violations number of exceedances of effluent TN over
N nitrogen (g m3)
18 gN m3 ()
NO nitrite plus nitrate nitrogen concentration
TN95 95th percentile for effluent TN (gN m3)
(gN m3)
TSS total suspended solids concentration (g l1)
OA,m mass percentage of oxygen in air (%)
TSS95 95th percentile for effluent TSS (g l1)
OA,v volume percentage of oxygen in air (%)
w mass air flow rate (kg s1)
OCI operational cost index (d1)
WWTP wastewater treatment plant
Oout volume percentage of oxygen in air leaving the
XBA autotrophic biomass concentration (gCOD m3)
surface of the aeration tank (%)
XBH heterotrophic biomass concentration (gCOD m3)
OTR oxygen transfer rate (g d1)
XI particulate inert organic material concentration
Patm atmospheric pressure (Pa)
(gCOD m3)
Pd pressure at the bottom of the aeration tank (Pa)
XND particulate biodegradable organic nitrogen
PE pumping energy (kWh d1)
concentration (gN m3)
PEeffluent contribution to pumping energy by effluent flow
XP particulate organic material concentration from
(kWh d1)
biomass decay (gCOD m3)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0 2183

XS particulate, slowly biodegradable organic material rA density of air at standard conditions (g m3)
concentration (gCOD m3) rsludge the density of sludge (kg m3)
y aerator depth (m) 4 temperature correction factor for oxygen transfer
a clean to process water correction factor () ()
b salinity-surface tension correction factor () u a factor exponent coefficient ()

strategies for conventional WWTPs in terms of effluent (FS) ones, to provide flexibility in membrane operation and
quality and operational costs, comprising a detailed descrip- cleaning (Itokawa et al., 2008), notwithstanding increased
tion of plant layout, models, input and evaluation criteria. pumping requirements. The characteristics and operation of
More recently, the importance of integrated control, plant- the membrane modules were based on commercially avail-
wide optimization and long-term evaluation was recognized able HF systems; minor modifications would be required for
within the wastewater treatment community and led to the a flat sheet configuration to be represented.
development of BSM1_LT (Rosen et al., 2004) and BSM2 All solids were assumed to be retained by the membrane.
(Jeppsson et al., 2006; Nopens et al., 2010). The widespread use Fouling of the membranes was not modelled as such, since no
of BSM, with more than 300 publications based on BSM1/2, consensus on its mechanisms has been reached. Coarse bubble
clearly indicates the usefulness of such a tool for the waste- aeration was incorporated in the model for fouling control so
water research community. that its impact on biology and operational costs, assuming
In this study, a dynamic benchmark simulation model for constant permeability, could be assessed. The design net flux
MBRs (BSM-MBR) is proposed as a platform to evaluate their was set to 20 l m2 h1 (LMH). Peak flows were assumed to incur
operational and control strategies. Control systems have a 100% increase in net flux to 40 LMH (Garcés et al., 2007).
already been proven for optimizing operational costs and Backwashing and relaxation were not physically modelled.
effluent quality for CAS plants (Olsson et al., 2005). The appli- 71500 m2 of membranes, divided over 8 separate 3.5m-high
cation of conventional control strategies for aeration, recircu- membrane tanks, were provided, enabling BSM-MBR to treat
lation pumping, carbon addition, etc. to MBRs is, however, yet the peak instantaneous storm flow with one membrane tank
to be thoroughly investigated. In terms of quantifying opera- out of service (worst-case scenario). 1500 m3 of membrane tank
tional costs for MBRs, thus far simple static spreadsheet volume was assumed to be required based on a packing density
models have been mainly adopted based on rules of thumb and of 47.5 m2 membrane area per m3 tank volume, which is at the
steady-state operation (Verrecht et al., 2008; Yoon et al., 2004). lower end of values reported in literature (Judd and Judd, 2010).
Although useful, these models may lead to erroneous conclu- A conservative specific membrane aeration demand (SADm) of
sions by not taking dynamic behavior and system configura- 0.3 Nm3 h1 per m2 of membrane area was chosen based on
tion into account, and precluding the evaluation of process literature values for hollow fibre systems (Judd and Judd, 2010),
control. These aspects can all be explored using BSM-MBR. resulting in a maximum of 21450 Nm3 h1 for coarse bubble
aeration of the membranes. The target membrane tank total
suspended solids (TSS ) concentration was 10 g l1.
2. Materials and methods
2.1.3. Tank sizing
BSM-MBR was given a total bioreactor volume of 7500 m3,
BSM-MBR is based on BSM1 (Alex et al., 2008; Copp, 2002). The
including the membrane tanks, resulting in an HRT of 3 h at
modification of BSM1 to provide BSM-MBR was conducted using
peak instantaneous storm flow and 9.8 h at average dry
the modelling and simulation software WEST (MOSTfor-
weather flow, which is within but at the lower end of values
WATER NV, Kortrijk, Belgium). Basic information on the BSM1/
reported for large MBRs in Europe (Itokawa et al., 2008).
BSM-MBR influent files is given in Table 1. For BSM-MBR, the
influent was assumed to already have passed pretreatment, i.e.
coarse screens, grit chamber, grease trap and fine sieves.

Table 1 e Flow-weighted average influent composition

2.1. Model configuration for BSM1 and BSM-MBR.
Compound Unit Dry Rain Storm
2.1.1. Biokinetics weather weather weather
As for BSM1, ASM1 (Henze et al., 2000) was used as biological
SI gCOD m3 30.00 25.96 28.03
process model for BSM-MBR. The BSM1 biokinetic parameter
SS gCOD m3 69.50 60.13 64.93
values were judged adequate for BSM-MBR; no consensus
XI gCOD m3 51.20 44.30 51.92
currently exists on updating the biokinetic values for MBRs XS gCOD m3 202.32 175.05 193.32
due to contradictory literature findings (Fenu et al., 2010a), XBH gCOD m3 28.17 24.37 27.25
and the default parameter values have been shown to be SNH gN m3 31.56 27.30 29.48
sufficient (Verrecht et al., 2010b). SND gN m3 6.95 6.01 6.49
XND gN m3 10.59 9.16 10.24
SALK molHCO 3 m
3
7.00 7.00 7.00
2.1.2. Membrane separation
Qi,av m3 d1 18446.33 21319.75 19744.72
Separate filtration tanks were used, as is common in almost
Qi,max m3 d1 32180.00 52126.00 60000.00
all submerged hollow fibre (HF) systems and many flat sheet
2184 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

Compared to BSM1 the total BSM-MBR volume was lower by The dissolved oxygen saturation concentration for clean water
37.5%, while the bioreactor volume was actually 25% higher. As at temperature T at sea level (C*(T ) e g m3) was calculated with
with BSM1, the total bioreactor volume was split into 5 zones: 2 the equation suggested by Benson and Krause (1984). The
anoxic zones followed by 3 aerobic zones, including the parameter values for Eq. (1) to (6) are given in Table 2. The chosen
membrane tanks. The anoxic volume fraction was set to 40%. values may be regarded as mean values, or at least within the
Thus, all zones were sized at 1500 m3. To accommodate range of cited literature values. Parameter values for a specific
a worst-case scenario of 25% of the bioreactor volume being out MBR system could differ from the values reported here. For open
of service, BSM-MBR was split up in 4 equal parallel lanes, as is loop operation (without control strategies implemented) a fine
actually the case for numerous full-scale WWTPs. As such, the bubble aeration flow of 6500 Nm3 h1 was selected, of which
actual volume of all 5m-high biological tanks was 375 m3. 4250 Nm3 h1 for the first aerobic zone and the remainder for the
second aerobic zone. The maximum possible fine bubble aeration
2.1.4. Sludge flows was set at 7000 Nm3 h1 per zone, based on manufacturer data.
To keep the sludge concentration in the membrane tanks The membrane tanks had no additional fine bubble aeration.
within reasonable limits and distribute it more evenly over the
whole plant, sludge was recirculated from the membrane
2.2. Evaluation criteria
tanks to the first aerobic zone at 55338 m3 d1, i.e. 3 times the
average DWF. Sludge was also recirculated from the second
The evaluation criteria of BSM1, these being the effluent
aerobic zone to the first anoxic zone at the same rate to recycle
quality index (EQI - kgPU d1) and the operational cost index
nitrate. Waste sludge was taken from the membrane tank
(OCI - d1), were used for BSM-MBR, with the latter adapted
recirculation loop (200 m3 d1) to maintain an SRT between 25
with reference to energy demand in kWh d1 from aeration
and 30 days as is common for MBRs (Itokawa et al., 2008). The
(AE ), pumping (PE ) and mixing (ME ).
general layout and flow scheme of BSM-MBR is shown in Fig. 1.
2.2.1. Aeration energy
2.1.5. Aeration The aeration energy for BSM-MBR was split into the contri-
1
In BSM1 the oxygen transfer rate (OTR - g d ) in the aerobic
butions from fine bubble aeration in the bioreactors
tanks is controlled by adapting the oxygen mass transfer
(AEbioreactor) and coarse bubble aeration in the membrane unit
coefficient. The aeration energy (AE e kWh d1) consumed is
(AEmembrane). Both were calculated by integration of the
calculated from this coefficient according to an empirical
expression for power requirement for adiabatic compression
formula. Using the equations of BSM1 for BSM-MBR would
(Tchobanoglous et al., 2003) over evaluation period Tev:
overlook the pivotal negative influence of elevated sludge
concentrations, which is paramount in MBR systems, on ZTev 
n  ZTev
24 wðtÞ,R,Tin pout 24
oxygen transfer efficiency (Henkel et al., 2009). For this reason, AE ¼ , , 1 ,dt ¼ ,cp , wðtÞ,dt (7)
Tev cSI ,n,e pin Tev
and to allow differentiation between coarse and fine bubble 0 0

aeration, a more fundamental and extensive aeration model

was adopted, combining several literature findings (Germain
et al., 2007; Judd and Judd, 2010; Stenstrom and Rosso, 2008; Table 2 e Oxygen transfer (top) and aeration energy
Tchobanoglous et al., 2003; Verrecht et al., 2008): (bottom) model parameter values.
Parameter Unit Value
OTR ¼ QA ,rA ,OA;m ,AOTE (1)

b 0.95
ðb,C;av ðTÞ  CÞ ðT20Þ F e 0.9a0.7b
AOTE ¼ SOTE,y, ,4 ,a,F (2) g m s2 9.81
C ð20Þ
OA,m % 23.2

 OA,v % 21
1  Pd Oout
C;av ðTÞ ¼ ,C ðTÞ, þ (3) Patm Pa 101325
2 Patm OA;v rA g m3 1200
rsludge kg m3 1000
a ¼ eu,TSS (4) SOTE % m1 2a6b

T C 15
y m 3.5a5b
Pd ¼ Patm þ rsludge ,g,h (5)
4 e 1.024
u e 0.05a0.083b
100,OA;v ,ð1  AOTEÞ
Oout ¼ (6)
ð1  OA;v ,AOTEÞ cSI e 29.7
e e 0.5
n e 0.283
pin Pa 101325
pout Pa 140660a155375b
R J mol1 K1 8.314
Tin K 293.15

a Coarse bubble aeration.

b Fine bubble aeration.
Fig. 1 e BSM-MBR layout and flow scheme.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0 2185

Eq. (7) combined with the parameter values in Table 2

Table 4 e Steady-state open loop BSM-MBR results for
provided a power consumption of 0.019 kWh Nm3 of air for reactor zones 1 to 5.
coarse bubble aeration and 0.025 kWh Nm3 for fine bubble
Compound Unit 1 2 3 4 5
aeration, comparable with literature values (Verrecht et al.,
3
2008). SI gCOD m 30.00 30.00 30.00 30.00 30.00
SS gCOD m3 2.25 1.31 0.85 0.77 0.67
XI gCOD m3 2678.62 2678.62 3554.43 3554.43 4722.18
2.2.2. Pumping energy
XS gCOD m3 82.52 76.19 65.13 59.35 67.25
As with BSM1, the PE for the sludge (PEsludge) was derived from
XBH gCOD m3 2699.15 2697.86 3573.19 3572.44 4739.59
three pumped sludge flows: the internal nitrate recirculation XBA gCOD m3 233.30 233.07 311.13 311.33 413.41
flow (Qint - m3 d1), the waste flow (Qw - m3 d1) and the return XP gCOD m3 1781.17 1782.50 2372.10 2373.11 3155.87
activated sludge flow (Qr - m3 d1). A value of 0.0075 kWh m3 SO g m3 0.01 0.00 2.46 2.19 8.11
was chosen for the pumping energy factors PF_Qint and PF_Qr, SNO gN m3 4.09 1.48 10.08 11.54 12.57
based on values for the MBR plants in Nordkanal and SNH gN m3 8.57 9.22 1.58 0.33 0.07
SND gN m3 1.08 0.68 0.65 0.63 0.58
Varsseveld (De Wever et al., 2009). The value of 0.05 kWh m3
XND gN m3 5.38 5.16 4.73 4.40 5.14
for PF_Qw was taken from BSM1.
SALK molHCO3 5.07 5.30 4.14 3.95 3.85
Pumping relating to effluent (or permeate) production m3
(PEeffluent) was calculated in the same way as for the sludge TSS g m3 5606.08 5601.19 7406.98 7403.00 9823.72
flows, with PF_Qe set to 0.075 kWh m3 based on values for
Nordkanal, Varsseveld (De Wever et al., 2009) and Schilde
(Fenu et al., 2010b). Changes in PF_Qe due to a varying filtration period, of dynamic dry, rain or storm weather. Closed and open
cycle, filtration flux and fouling behavior were ignored. loop results refer to BSM-MBR simulations respectively with
and without control strategies implemented. The full
2.2.3. Mixing energy membrane and biological capacity was used in all simulations.
The total mixing energy (ME - kWh d1) comprised the energy
used for mixing the anoxic, aerobic tanks and membrane
tanks. The anoxic tanks were mixed constantly and required 3. Results and discussion
0.008 kW mixing power per m3 tank volume (Fenu et al., 2010b;
Tchobanoglous et al., 2003), yielding a constant ME of 576 kWh 3.1. Steady-state open loop evaluation
d1. The threshold value for sufficient aeration for mixing was
set at 165 Nm3 h1 for an aerobic tank and 120 Nm3 h1 for The steady-state results for the open loop case of BSM-MBR
a membrane tank, based on a value of 2.2 m3 h1 per m2 are shown in Tables 3 and 4. The results in Table 3 show that
ground surface area (Water Environment Federation, 2009). BSM-MBR performs better in terms of effluent TSS and COD
Below the threshold additional mechanical mixing at compared to BSM1, mostly because of the full retention of
0.008 kW m3 was assumed necessary. particulates by the membranes. In terms of N removal, it can
be observed that superior nitrification is obtained in BSM-
2.3. Simulation procedure MBR. Effluent nitrate concentrations are however higher for
BSM-MBR than BSM1 due to excessive aeration providing
Steady-state and dynamic simulations with BSM-MBR were complete nitrification, while influent carbon for denitrifica-
performed in the same way as described for BSM1, i.e. steady- tion is limited with a COD to TN ratio of only 6.93 gCOD gN1.
state simulation up to 10 times the sludge age followed by three Moreover, less nitrate and more oxygen is recycled back to the
weeks of dynamic dry weather and a last week, the evaluation anoxic zone of BSM-MBR compared to BSM1, causing
a reduced denitrification performance. The high DO levels in
zones 3, 4 and 5 in Table 4 also indicate inhibited simulta-
neous nitrification-denitrification. The total SRT for BSM-MBR
Table 3 e Comparison of BSM-MBR and BSM1 steady- is 27.4 days, which is within the intended limits. The anoxic
state open loop effluent results.
mass fraction amounts to only 31%, despite an anoxic volume
Compound Unit BSM1 BSM-MBR fraction of 40%, due to the steep TSS gradient along the
SI gCOD m3 30.00 30.00 different reactor zones: 5.6 g l1, 7.4 g l1 and 9.8 g l1 for the
SS gCOD m3 0.89 0.67 anoxic, aerobic and membrane tanks respectively.
XI gCOD m3 4.39 0.00
XS gCOD m3 0.19 0.00
3.2. Dynamic open loop evaluation
XBH gCOD m3 9.78 0.00
XBA gCOD m3 0.57 0.00
XP gCOD m3 1.73 0.00 The dynamic dry, rain and storm weather results for the open
SO g m3 0.49 8.11 loop case of BSM-MBR are shown in Tables 5 and 6. From Table
SNO gN m3 10.42 12.57 5 it is clear that the use of membrane filtration instead of
SNH gN m3 1.73 0.07 secondary clarification ameliorates adverse effects of rain and
SND gN m3 0.69 0.58 storm weather conditions on effluent quality, since sludge
XND gN m3 0.01 0.00
wash-out is not possible. The dynamic dry weather results in
SALK molHCO3 m
3
4.13 3.85
Table 5 are comparable to the steady-state results in Table 3
TSS g m3 12.50 0.00
for BSM-MBR, but not for BSM1 in terms of SNH and SNO.
2186 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

Table 5 e Comparison of BSM-MBR and BSM1 dynamic open loop flow proportionally averaged effluent results for dry, rain
and storm weather.
Compound Unit BSM1 BSM-MBR

Dry Rain Storm Dry Rain Storm

3
SI gCOD m 30.00 22.84 26.30 30.00 22.85 26.29
SS gCOD m3 0.97 1.13 1.11 0.70 0.72 0.74
XI gCOD m3 4.58 5.64 5.64 0.00 0.00 0.00
XS gCOD m3 0.22 0.34 0.32 0.00 0.00 0.00
XBH gCOD m3 10.22 12.86 11.88 0.00 0.00 0.00
XBA gCOD m3 0.54 0.64 0.59 0.00 0.00 0.00
XP gCOD m3 1.76 2.07 1.91 0.00 0.00 0.00
SO g m3 0.75 0.85 0.76 7.58 7.00 6.99
SNO gN m3 8.82 6.96 7.48 12.74 11.20 11.78
SNH gN m3 4.76 4.98 5.35 0.12 0.12 0.13
SND gN m3 0.73 0.82 0.80 0.60 0.61 0.62
XND gN m3 0.02 0.02 0.02 0.00 0.00 0.00
SALK molHCO3 m
3
4.46 5.14 4.87 3.85 4.49 4.19
TSS g m3 12.99 16.16 15.26 0.00 0.00 0.00
TKN gN m3 6.75 7.37 7.63 0.72 0.74 0.76
TN gN m3 15.57 14.32 15.11 13.46 11.93 12.54
COD g m3 48.30 45.52 47.76 30.70 23.58 27.03
BOD5 g m3 2.77 3.47 3.23 0.18 0.18 0.18
Qe m3 d1 18061.33 23808.19 20658.08 18246.31 23993.17 20843.08

Apparently the nitrification capacity of BSM1 is at times first and second aerobic zone are stable (not shown). Having
insufficient during dynamic simulations. BSM-MBR has 12.5% constant internal nitrate recirculation and return activated
more aerobic volume compared to BSM1 and also carries more sludge flows is clearly insufficient for maintaining a stable
than two times the biological mass per unit volume. The sludge distribution over the plant at all times. The combina-
excessive membrane aeration in BSM-MBR further ensures DO tion of a higher demand for oxygen during peak flows and less
levels sufficiently high to maintain nitrification capacity efficient aeration at high TSS induces high variability in the DO
during dynamic conditions. levels of the membrane tanks during dynamic simulations.
The impact of influent dynamics on TSS and DO concen- The DO in the other aerobic zones is also highly variable. Even
trations throughout BSM-MBR is clearly visible in Fig. 2. With under normal dry weather conditions the DO in the second
every peak flow sludge is washed out the anoxic tanks aerobic zone fluctuates from 0.25 mg l1 to 6 mg l1. The
towards the membrane tanks. The TSS concentrations in the former has, as mentioned before, little effect on effluent

Table 6 e Comparison of BSM-MBR and BSM1 dynamic open loop effluent quality and operational cost performance criteria
for dry, rain and storm weather.
Criterion Unit BSM1 BSM-MBR

Dry Rain Storm Dry Rain Storm

1
IQI kgPU d 52081.40 52081.40 54061.50 52081.40 52081.40 54061.50
EQI kgPU d1 6690.73 8936.23 8022.77 3286.54 3790.07 3499.88
TN95 gN m3 18.54 17.79 18.72 16.83 15.75 16.74
SNH,95 gN m3 8.88 9.47 9.78 0.37 0.37 0.38
TSS95 g m3 15.75 21.69 20.79 0.00 0.00 0.00
TNlimit_violations e 5 3 4 0 0 0
(18 gN m3) % of time 8.18 4.32 8.48 0.00 0.00 0.00
SNH,limit_violations e 7 7 7 0 0 0
(4 gN m3) % of time 62.50 63.24 64.43 0.00 0.00 0.00
SPtotal kgTSS d1 2670.32 2737.14 2914.53 1961.12 1974.90 2166.26
e SP kgTSS d1 2435.67 2352.32 2599.36 1961.12 1974.90 2166.26
AE kWh d1 3341.39 3341.39 3341.39 13558.87 13558.87 13558.87
e AEbioreactor kWh d1 3341.39 3341.39 3341.39 3878.45 3878.45 3878.45
e AEmembrane kWh d1 e e e 9680.42 9680.42 9680.42
PE kWh d1 388.17 388.17 388.17 2208.54 2639.56 2403.30
e PEsludge kWh d1 388.17 388.17 388.17 840.07 840.07 840.07
e PEeffluent kWh d1 e e e 1368.47 1799.49 1563.23
ME kWh d1 240.00 240.00 240.00 576.00 576.00 576.00
OCI d1 16147.92 15731.18 16966.34 26148.99 26648.91 27369.45
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0 2187

Fig. 2 e Impact of dry, rain and storm weather influent dynamics on TSS and DO in the membrane tanks, DO in the second
aerobic zone and TSS in the first anoxic zone. The 2nd and 3rd day of the 7 day evaluation period are shown.

ammonium concentrations because of the excessive differ significantly. However, aeration costs can be expected to
membrane aeration, the latter causes severe oxygen be higher for MBRs than CAS plants. 71% of the aeration costs
poisoning of the first anoxic zone. for BSM-MBR are linked with the coarse bubble aeration for
BSM-MBR performs 51% (dry weather), 58% (rain weather) membrane fouling control, while it was calculated that the
and 56% (storm weather) better than BSM1 in terms of EQI latter accounts for only 30e31% of oxygen transferred into
(Table 6), and no effluent limits are violated. Nonetheless, the system with 2e3% of the total oxygen lost through the
BSM-MBR effluent TN can be high at times (as indicated by effluent. The elevated pumping energy costs for BSM-MBR can
TN95) due to poor denitrification (as indicated by SNH,95). mostly be attributed to permeate production through
Compared to the dry weather situation, the BSM-MBR EQI membrane filtration. Also, more sludge is being pumped
increases 15% and 6% for the rain and storm weather case
respectively, whereas the corresponding BSM1 EQI figures are
34% and 20%. BSM-MBR is thus more stable than BSM1 when Table 7 e Overview of total and specific energy costs for
subjected to varying influent conditions. the MBRs of Schilde (Fenu et al., 2010b), Varsseveld (De
However, the superior effluent quality of BSM-MBR incurs Wever et al., 2009), Nordkanal (Brepols et al., 2010) and
BSM-MBR under dry weather conditions.
a cost 61e69% higher than that of BSM1 depending on influent
dynamics, according to the OCI. Other than for sludge Energy cost Schilde Varsseveld Nordkanal BSM-MBR
(kWh m3)
disposal, all costs are increased significantly (140% for mixing,
306% for aeration and up to 580% for pumping). The higher ME 0.05 0.04 0.11 0.03
mixing costs can be attributed to the larger anoxic volume to PEsludge 0.10 0.11 0.01 0.05
be mixed and the higher energy factor for mixing selected to PEeffluent 0.07 0.12 0.02 0.07
AEbioreactor 0.07 0.24 0.11 0.21
incorporate the influence of elevated TSS on mixing. Care
AEmembrane 0.23 0.34 0.45 0.53
should be taken when comparing aeration costs between
Total 0.52 0.85 0.71 0.90
BSM1 and BSM-MBR, since their respective aeration models
2188 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

Table 8 e BSM-MBR dynamic closed loop effluent quality and operational cost performance criteria for dry, rain and storm
weather.
Criterion Unit DO control DO þ SADm control

Dry Rain Storm Dry Rain Storm

1
IQI kgPU d 52081.40 52081.40 54061.50 52081.40 52081.40 54061.50
EQI kgPU d1 3224.22 3717.95 3461.05 3203.27 3702.66 3440.81
TN95 gN m3 17.46 16.21 17.25 17.37 16.13 17.17
SNH,95 gN m3 0.17 0.18 0.17 0.17 0.18 0.18
TNlimit_violations e 4 1 4 4 1 4
(18 gN m3) % of time 0.03 0.01 0.03 0.02 0.01 0.02
SP kgTSS d1 1961.17 1975.06 2166.11 1961.20 1975.07 2166.14
AE kWh d1 13142.86 13106.08 13234.11 9122.00 10120.26 9584.75
e AEbioreactor kWh d1 3462.44 3425.66 3553.69 3525.63 3471.58 3612.54
e AEmembrane kWh d1 9680.42 9680.42 9680.42 5596.38 6648.68 5972.21
PE kWh d1 2208.55 2639.56 2403.30 2208.55 2639.56 2403.30
e PEsludge kWh d1 840.07 840.07 840.07 840.07 840.07 840.07
e PEeffluent kWh d1 1368.48 1799.49 1563.23 1368.48 1799.49 1563.23
ME kWh d1 576.00 576.00 576.00 576.00 576.00 576.00
OCI d1 25733.28 26196.96 27043.97 21712.54 23211.16 23394.74

around in BSM-MBR than BSM1. The significant decrease in 0.002 d. The results in Table 8 show the proposed DO control
sludge production for disposal, 16e19%, can be explained by strategy impact on effluent quality being marginally beneficial,
the more than three times longer SRT of BSM-MBR compared if not the contrary, compared to the open loop case, with EQI
to BSM1 (Lubello et al., 2009). decreasing 1e2% depending on the weather conditions, but the
TN effluent limit also being violated in each case. The cost of
3.3. Comparison with full-scale MBRs fine bubble aeration decreased significantly, 8e12% compared
with the open loop case, albeit with only a minor impact on
The total specific energy requirement of modern, optimized overall OCI since the latter is dominated by sludge disposal and
large-scale MBR plants is reported as being in the range membrane aeration costs.
0.6e1 kWh m3 (Lesjean et al., 2009). Table 7 provides
a breakdown of energy costs per m3 of permeate for three 3.4.2. DO and SADm control
large-scale MBR plants (Schilde, Varsseveld and Nordkanal) Extending the former control scheme to link membrane aera-
compared with the dry weather open loop results of BSM-MBR. tion to flux, assuming this to have no major adverse effects on
Notwithstanding some energy costs being very plant specific, membrane fouling and sustainable flux (Garcés et al., 2007;
it seems that the BSM-MBR energy costs are comparable with Stone and Livingston, 2008), was tested. SADm was assumed to
those from full-scale plants. Only membrane aeration costs decrease linearly from 0.3 to 0.15 Nm3 h1 m2 with fluxes
are consistently higher for BSM-MBR, since the membrane decreasing from 20 to 10 LMH. Beyond these limits SADm
aeration was constantly applied to all membranes in the open remained constant. Again, sensor and actuator performance
loop simulations for BSM-MBR, whereas in reality membrane was assumed ideal. The results in Table 8 show the SADm
tanks are taken in and out of service depending on influent control scheme to have a minor effect on effluent quality, with
flow and membrane flux. The MBRs of Schilde, Varsseveld and EQI decreasing by 0e1% compared to the closed loop case with
Nordkanal are to some extent optimized, which BSM-MBR in only DO control. The membrane aeration costs decrease by 42,
its open loop form by definition is not. 31 and 38% for the dry, rain and storm case respectively, while
the fine bubble aeration costs increase marginally, i.e. 1e2%, to
3.4. Closed loop performance satisfy biological oxygen demand. Interestingly, diminishing
membrane aeration has only minor effect on oxygen transfer
The impact of imposing a basic control and novel operational since the latter still accounts for 27e29% of the oxygen trans-
strategy for regulating aeration was studied for illustrative ferred to the system (results not shown). The explanation lies in
purposes. the lower DO levels obtained in the membrane tanks when
membrane aeration is lowered. This increases the driving force
3.4.1. DO control for oxygen transfer, while, depending on the weather condi-
An aeration control scheme was implemented maintaining the tions, also 15e24% less oxygen is lost through the effluent.
DO concentration in the second aerobic zone at 1.5 mg l1 using Compared to the open loop case, the overall OCI decreases by
a PI controller to adjust the fine bubble aeration in both the first 13e17%. The results thus show large potential for saving energy
and second aerobic zone. Moreover, 50% more air was sent to by having proportional membrane aeration without compro-
the first than the second aerobic zone, since it receives a higher mising effluent quality. The latter may, however, be compro-
load, unless the maximum aeration capacity has been reached. mised when proportional membrane aeration is used combined
The DO sensor and actuator performance was assumed to be with other operational and control strategies. Also, a thorough
ideal, i.e. without noise or delay. The proportional gain of the investigation of the technical feasibility and fouling control
controller was tuned to 500 Nm3 h1 and the integral time to effectiveness of proportional membrane aeration is needed.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0 2189

bioreactors for municipal applications. Water Science and

4. Conclusions Technology 61 (10), 2461e2468.
Copp, J.B., 2002. The COST Simulation Benchmark - Description
 A benchmark simulation model for MBRs (BSM-MBR) has and Simulator Manual. Office for Official Publications of the
been developed. The existing BSM1 for a conventional European Communities, Luxembourg.
De Wever, H., Brepols, C., Lesjean, B., 2009. Decision Tree for Full-
WWTP was used as starting point and updated in terms of
scale Submerged MBR Configurations. Final MBR-Network
reactor volumes, membrane filtration, aeration capacity and
Workshop, 31 March - 1 April 2009 (Berlin, Germany).
sludge flows. The BSM1 performance criteria were extended Fenu, A., Guglielmi, G., Jimenez, J., Sperandio, M., Saroj, D.,
for an MBR taking into account additional pumping Lesjean, B., Brepols, C., Thoeye, C., Nopens, I., 2010a. Activated
requirements for permeate production and aeration sludge model (ASM) based modelling of membrane bioreactor
requirements for fouling suppression. A dedicated aeration (MBR) processes: a critical review with special regard to MBR
model was used to incorporate the effects of elevated sludge specificities. Water Research 44 (15), 4272e4294.
Fenu, A., Roels, J., Wambecq, T., De Gussem, K., Thoeye, C., De
concentrations on aeration efficiency and costs.
Gueldre, G., Van De Steene, B., 2010b. Energy audit of a full
 Steady-state and dynamic open loop simulations revealed scale MBR system. Desalination 262 (1e3), 121e128.
the effluent quality of BSM-MBR to be up to 58% better than Garcés, A., De Wilde, W., Thoeye, C., De Gueldre, G., 2007.
that of BSM1, mainly thanks to the complete retention of Operational Cost Optimisation of MBR Schilde. 4th IWA
solids and improved ammonium removal due to extensive International Membranes Conference: Membranes for Water
aeration in combination with more biological mass. and Wastewater Treatment, 15e17 May 2007 (Harrogate, UK).
However, this was at the expense of significantly higher Germain, E., Nelles, F., Drews, A., Pearce, R., Kraume, M., Reid, E.,
Judd, S.J., Stephenson, T., 2007. Biomass effects on oxygen transfer
operational costs. Only the sludge disposal costs decreased
in membrane bioreactors. Water Research 41 (5), 1038e1044.
for the BSM-MBR, due to the higher SRT. Gernaey, K.V., van Loosdrecht, M.C.M., Henze, M., Lind, M.,
 Impaired denitrification performance was evident due to Jorgensen, S.B., 2004. Activated sludge wastewater treatment
oxygen poisoning of the first anoxic zone and a reduced plant modelling and simulation: state of the art.
anoxic mass fraction related to the steep TSS gradient along Environmental Modelling & Software 19 (9), 763e783.
the bioreactor zones. Furthermore, the TSS gradient was Henkel, J., Lemac, M., Wagner, M., Cornel, P., 2009. Oxygen
transfer in membrane bioreactors treating synthetic
found to be highly susceptible to influent flow dynamics,
greywater. Water Research 43 (6), 1711e1719.
also having repercussions on aeration efficiency.
Henze, M., Gujer, W., Mino, T., van Loosdrecht, M., 2000. Activated
 A comparison with three large-scale MBRs showed BSM- Sludge Models: ASM1, ASM2, ASM2d and ASM3. IWA
MBR energy costs to be realistic. The membrane aeration Publishing, London.
costs for the open loop simulations were high due to the lack Itokawa, H., Thiemig, C., Pinnekamp, J., 2008. Design and
of optimization. operating experiences of municipal MBRs in Europe. Water
 Two closed loop simulations were run to show the potential Science and Technology 58 (12), 2319e2327.
Jeppsson, U., Pons, M.N., 2004. The COST benchmark simulation
of control strategies applied to BSM-MBR for diminishing
model - current state and future perspective. Control
operational costs by 13e17% depending on influent Engineering Practice 12 (3), 299e304.
dynamics, without compromising effluent quality. Jeppsson, U., Rosen, C., Alex, J., Copp, J., Gernaey, K., Pons, M.N.,
Vanrolleghem, P.A., 2006. Towards a benchmark simulation
model for plant-wide control strategy performance evaluation
of WWTPs. Water Science and Technology 53 (1), 287e295.
Judd, S., 2008. The status of membrane bioreactor technology.
Acknowledgements Trends in Biotechnology 26 (2), 109e116.
Judd, S.J., Judd, C., 2010. The MBR Book - Second Edition:
Principles and Applications of Membrane Bioreactors in Water
The authors want to thank Christoph Brepols (Erftverband) for
and Wastewater Treatment. Elsevier, London, UK.
his input on MBR costs and control strategies and Lorenzo Kinnear, D.J., Pellegrin, M.-L., Cross, T.B., Condran, M.J.,
Benedetti and Phuong Thu Pham (BIOMATH) for their assis- Kochaba, T., Haney, C.M., 2010. Comparing Membrane
tance in modelling BSM-MBR. Thomas Maere is supported by Bioreactors and Conventional Activated Sludge Processes for
the Institute for Encouragement of Innovation by means of Low Nutrient Limits Membrane Applications 2010, 6e9 June
Science and Technology in Flanders (IWT). 2010, Anaheim, California.
Lesjean, B., Ferre, V., Vonghia, E., Moeslang, H., 2009. Market and
design considerations of the 37 larger MBR plants in Europe.
references
Desalination and Water Treatment 6 (1e3), 227e233.
Lubello, C., Caffaz, S., Gori, R., Munz, G., 2009. A modified Activated
Sludge Model to estimate solids production at low and high
Alex, J., Benedetti, L., Copp, J., Gernaey, K.V., Jeppsson, U., Nopens, I., solids retention time. Water Research 43 (18), 4539e4548.
Pons, M.-N., Rieger, L., Rosen, C., Steyer, J.P., Vanrolleghem, P., Nopens, I., Benedetti, L., Jeppsson, U., Pons, M.N., Alex, J., Copp, J.B.,
Winkler, S., 2008. Benchmark Simulation Model No. 1 (BSM1). Gernaey, K.V., Rosen, C., Steyer, J.P., Vanrolleghem, P.A., 2010.
Division of Industrial Electrical Engineering and Automation, Benchmark Simulation Model No 2: finalisation of plant layout
Lund University. http://www.benchmarkwwtp.org/. and default control strategy. Water Science and Technology 62
Benson, B.B., Krause, D., 1984. The concentration and Isotopic (9), 1967e1974.
Fractionation of oxygen dissolved in Fresh-water and Olsson, G., Nielsen, M.K., Yuan, Z., Lynggaard-Jensen, A., Steyer, J.P.,
Seawater in Equilibrium with the Atmosphere. Limnology and 2005. Instrumentation, Control and Automation in Wastewater
Oceanography 29 (3), 620e632. Systems. IWA Publishing, London.
Brepols, C., Schafer, H., Engelhardt, N., 2010. Considerations on Rosen, C., Jeppsson, U., Vanrolleghem, P.A., 2004. Towards
the design and financial feasibility of full-scale membrane a common benchmark for long-term process control and
2190 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 2 1 8 1 e2 1 9 0

monitoring performance evaluation. Water Science and Verrecht, B., Maere, T., Benedetti, L., Nopens, I., Judd, S., 2010a.
Technology 50 (11), 41e49. Model-based energy optimisation of a small-scale
Stenstrom, M.K., Rosso, D., 2008. Aeration and mixing. In: decentralised membrane bioreactor for urban reuse. Water
Henze, M., van Loosdrecht, M., Ekama, G.A., Brepols, C. (Eds.), Research 44 (14), 4047e4056.
Biological Wastewater Treatment: Principles, Modelling and Verrecht, B., Maere, T., Nopens, I., Brepols, C., Judd, S., 2010b. The
Design. IWA Publishing, London. cost of a large-scale hollow fibre MBR. Water Research 44 (18),
Stone, M., Livingston, D., 2008. Flat Plate MBR Energy Consumption 5274e5283.
- Village of Dundee, MI. WEF Membrane Technology Conference Water Environment Federation, 2009. Energy Conservation in
27e30 January 2008, Alexandria, Virginia. Water and Wastewater Facilities. McGraw-Hill
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Professional.
Engineering: Treatment and Reuse. McGraw-Hill, Boston. Yoon, S.H., Kim, H.S., Yeom, I.T., 2004. The optimum
Verrecht, B., Judd, S., Guglielmi, G., Brepols, C., Mulder, J.W., operational condition of membrane bioreactor (MBR): cost
2008. An aeration energy model for an immersed estimation of aeration and sludge treatment. Water Research
membrane bioreactor. Water Research 42 (19), 4761e4770. 38 (1), 37e46.

View publication stats