Content uploaded by Ramon Vilanova
Author content
All content in this area was uploaded by Ramon Vilanova on Jun 02, 2021
Content may be subject to copyright.
Manipulating internal recirculation flow rate on the
biological process in wastewater treatment
I. Sant´
ın, R. Vilanova, C. Pedret
Dept. Telecommunication and Systems Engineering
Autonomous University of Barcelona
Bellaterra (Barcelona), Spain
{Ignacio.Santin, Ramon.Vilanova, Carles.Pedret}@uab.cat
M. Barbu
Automatic Control and Electrical Engineering Dept.
”Dunarea de Jos” Univerity of Galati
Galati, Romania
Marian.Barbu@ugal.ro
Abstract—Internal recirculation plays an important role in the
different biological treatment areas of a wastewater treatment
plant, because it has a great influence on the concentration
of pollutants, especially nutrients. The most common uses of
the internal recirculation flow rate are to keep it fixed or to
manipulate it to obtain a fixed nitrate set-point in the last anoxic
tank using control techniques. This work proposes a new control
strategy for manipulating the internal recirculation flow rate
by applying a fuzzy controller. The proposed controller takes
into account the effects of the internal recirculation flow rate
at the inlet of the biological treatment and in the denitrification
and nitrification processes with the aim of reducing violations
of legally established limits of nitrogen and ammonia and also
reducing operational costs. The proposed fuzzy controller is
tested by simulation with the internationally known benchmark
simulation model no. 2. The default control strategy with a
hierarchical dissolved oxygen control is used and the results
are compared by keeping the recirculation flow fixed (as the
default control strategy) and manipulating it with the proposed
fuzzy controller . The results show reduction improvements of
68.78% of total nitrogen limit violations, 24.92% of ammonia
limit violations and 3.79% of pumping energy costs.
Index Terms—wastewater treatment plant, fuzzy control,
benchmark simulation model no. 2, control strategies
I. INTRODUCTION
Large amounts of freshwater are used daily around the
world, becoming waste water. This waste water must be
treated to avoid contamination of the receiving waters (rivers,
lakes, etc.), which could affect aquatic life and consequently
biodiversity. Due to this reason, wastewater treatment plants
(WWTPs) are necessary to maintain the required levels of
water quality.
Specifically, maximum concentration limits are established
for discharges in the receiving environment of Total Suspended
Solids (TSS), organic matter (Biological Oxygen Demand in 5
days (BOD5), and Chemical Oxygen Demand (COD)), total
nitrogen (SNtot), phosphorous and ammonium (SNH ). Nitrogen
and phosphorus are nutrients that can cause eutrophication
in the receiving water, and consequently the death of aquatic
beings. SNH, in addition to containing nitrogen, is toxic to
aquatic life. Precisely, keeping SNtot and SNH concentrations
below the limits is usually one of the most difficult objectives
to fulfill in WWTPs. In order to achieve all these targets, the
application of control strategies in WWTPs are very common.
Given the importance of keeping the pollutant concentration
within the established limits, and achieving this with the
lowest possible operational costs, several research works have
been published in recent years focusing on the application
of control strategies in WWTPs. Some works apply control
strategies in the sludge treatment as in [1] and [2], but most
articles do it in the secondary treatment, which corresponds to
biological treatment, whose operation is explained in Section
II. In the literature there are several works that aim to improve
the control of the concentration of the dissolved oxygen
(SO) in the aerobic reactors, by manipulating the oxygen
transfer coefficient (KLa), using different control techniques
( [3]–[7]). Improving the regulation of the SOset-point in the
nitrification process has also been the goal of many works,
such as [8]–[13]. Another variable that can be manipulated
in the biological treatment is the internal recirculation flow
rate (Qa). However, research into a new control strategy to
manipulate Qais not common. In some control strategies of
the works cited above, Qais kept fixed and in others Qais
manipulated to maintain the nitrate (SNO) set-point in the last
anoxic reactor. For example, in [14], the usual manipulation
of Qais modified to avoid SNH violations, but only in the
periods of time when a risk of SNH violation is predicted.
The novelty proposed in this paper is a new control strategy
to manipulate Qaby a fuzzy controller. The variation of
Qahas several effects on the nitrification and denitrification
processes of the biological treatment, and this paper argues
why the proposed control strategy is more beneficial than
the usual ones, on the basis of these effects. For the design
of the proposed fuzzy controller, the dilution or the increase
of concentration of different compounds at the beginning of
the biological treatment is taken into account, as well as the
variation of the Hydraulic Retention Time (HRT) (a more
exhaustive explanation is found in Section III. This paper
uses an already published control strategy as the baseline
operational control strategy, replacing only the proposed Qa
manipulation, and shows the improvements it provides in
terms of reducing violations of SNtot and SNH, as well as
operational costs.
2020 24th International Conference on System Theory, Control and Computing (ICSTCC)
978-1-7281-9809-5/20/$31.00 ©2020 IEEE 588
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
The use of simulation models to test control strategies
in WWTPs is a common practice in research. So much
so that most of the previously referenced papers use the
internationally known and accepted standard Benchmark
Simulation Model no. 2 (BSM2) ( [15]), which is an
extension of the Benchmark Simulation Model no. 1 (BSM1)
developed by the International Association on Water Pollution
Research and Control ( [16], [17]). BSM2 differs from BSM1
by including the entire cycle of a WWTP, adding the sludge
treatment, and in that the simulation period is extended
to one-year assessment. In this work, the simulations and
evaluations of the control strategies have been carried out
with BSM2.
The paper is organised as follows. First, BSM2 is presented.
Next, the proposed fuzzy control is explained. Afterward,
the simulation results are shown, as well as the discussion
about them. Finally, the most important conclusions are drawn.
II. MATERIALS AND METHODS
The evaluation of the proposed fuzzy controller has been
carried out using the internationally known BSM2 ( [18])
which was updated by [19]. The BSM2 layout (Fig. 1) includes
the biological treatment (secondary treatment) of BSM1,
adding a primary settler and sludge treatment by a thickener,
a digester and a dewatering. The liquid resulting from the
latter is stored in a tank and recirculated to the primary settler.
Fig. 1. BSM2 plant with notation used for flow rates
BSM2 includes dynamics of the different influent
concentrations for a period of 609 days, but only the data
from day 245 to 609 (one year) are evaluated. These dynamics
include rain events and temperature (Tas) variations.
The biological treatment consists of five activated sludge
reactors, followed by a secondary settler. The first two
reactors are anoxic, where the denitrification process is
carried out and the following three tanks are aerobics, where
the nitrification process takes place. The influent has an
average dry weather flow rate of 20648.36 m3/d and an
average biodegradable chemical oxygen demand (COD) of
592.53 mg/l. The volume of each anoxic tank is 1500 m3and
of each aerobic tank is 3000 m3. The hydraulic retention time
of the biological treatment is 14 hours. There is an internal
recirculation from the last tank to the first one and another
from the underflow of the settler in order to recirculate sludge.
The Activated Sludge Model No. 1 (ASM1) [20] describes
the processes of the biological reactors. They define the
conversion rates of the different variables of the biological
treatment. The present work is based on the conversion
rates of of SNH (rSNH ) and SNO (rSNO ) for the design of the
proposed fuzzy controller taking into account the Qaeffects
on the biological treatment. They are shown following:
rNH =0.08⇢10.08⇢2✓0.08 + 1
0.24◆⇢3+⇢6(1)
rNO =0.1722⇢2+4.1667⇢3(2)
where ⇢1,⇢2,⇢3,⇢6are four of the eight biological processes
defined in ASM1. Specifically, ⇢1is the aerobic growth of
heterotrophs, ⇢2is the anoxic growth of heterotrophs, ⇢3is
the aerobic growth of autotrophs and ⇢6is the ammonification
of soluble organic nitrogen. They are defined below:
⇢1=µHT ✓SS
10 + SS◆✓ SO
0.2+SO◆XB,H (3)
where SSis the readily biodegradable substrate and µHT is:
µHT =4·exp Ln 4
3
5!·(Tas 15)!(4)
⇢2=µHT ✓SS
10 + SS◆✓ 0.2
0.2+SO◆✓ SNO
0.5+SNO ◆0.8·XB,H
(5)
⇢3=µAT ✓SNH
1+SNH ◆✓ SO
0.4+SO◆XB,A (6)
where XB,A is the active autotrophic biomass and µAT is:
µAT =0.5·exp Ln 0.5
0.3
5!·(Tas 15)!(7)
⇢6=kaT ·SND ·XB,H (8)
where SND is the soluble biodegradable organic nitrogen and
kaT is:
2
589
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
kaT =0.05 ·exp Ln 0.05
0.04
5!·(Tas 15)!(9)
The general equations for mass balancing are:
•For reactor 1:
dZ1
dt =1
V1
(Qa·Za+Qr·Zr+Qpo ·Zpo +rz,1·V1Q1·Z1)
(10)
where Z is any concentration of the process, Z1is Z in
the first reactor, Zais Z in the internal recirculation, Zris
Z in the external recirculation, Zpo is Z from the primary
clarifier, V is the volume, V1is V in the first reactor, Qpo
is the overflow of the primary clarifier and Q1is the flow
rate in the first tank and it is equal to the sum of Qa,Q
r
and Qpo.
•For reactor 2 to 5:
dZk
dt =1
Vk
(Qk1·Zk1+rz,k ·VkQk·Zk)(11)
where k is the number of reactor and Qkis equal to Qk-1
The results evaluation is carried out by the effluent quality
and the operational costs. Effluent quality is evaluated
by the percentage of time that the effluent concentrations
of SNtot, total COD, SNH , TSS and BOD5are above the
established limits, shown in Table I. Costs are evaluated by
the Overall Cost Index (OCI) and with each of its components.
TABLE I
EFFLU ENT Q UAL ITY L IM ITS
Variable Value
SNtot <18 g N.m3
CODt<100 g COD.m3
SNH <4 g N.m3
TSS <30 g SS.m3
BOD5<10 g BOD.m3
OCI is defined to evaluate the operational cost as:
OCI =AE+PE+3·SP+3·EC+ME6·METprod+HEnet
(12)
where AE is the aeration energy, PE is the pumping energy,
SP is the sludge production to be disposed, EC is the
consumption of external carbon source, ME is the mixing
energy, METprod is the methane production in the anaerobic
digester and HEnet is the net heating energy.
III. CONTROL APPROACH
This article applies a new control strategy to manipulate
Qaby a fuzzy controller, with the aim of reducing SNtot in the
effluent (SNtot,e) and SNH in the effluent (SNH,e) violations as
well as operational costs. The idea of the work is to add the
proposed fuzzy controller to an already tested and published
control strategy, replacing only the Qamanipulation. The
effects of Qaon biological treatment, the design of the
proposed fuzzy controller and the control strategy on which
it is tested are detailed below.
A. Qaeffects on the biological treatment
The Qavariation influences both the denitrification and the
nitrification processes, as well as the dilution or increase of
concentrations at the biological treatment input. Due to this
fact Qavariations have immediate effects on SNtot,e and SNH,e,
but also other different effects on the same variables after
some time. This complexity makes necessary an in-depth
knowledge of the plant behavior for the Qamanipulation and
justify the use of a fuzzy controller.
A more detailed explanation of the Qainfluence on the
biological treatment, specifically on SNtot and SNH, is carried
out by the equations of the observed conversion rates, mass
balance of the biological reactors described in Section II and
the mixture of concentrations at the inlet of the biological
treatment.
1) Qaeffect at the inlet of the biological treatment: On one
hand, the SNH value at the inlet of the first reactor is given
by the mixture of SNH from the primary treatment (SNH,po)
and the recirculated SNH (which is SNH in the fifth reactor
(SNH,5)), as can be seen in (13). As due to the nitrification
process the SNH,5 value is much less than SNH,po,Qaincreases
causes SNH,0 dilution and, on the contrary, Qareductions
result in SNH,0 increases. On the other hand, since there is no
SNO in the influent and all SNO at the inlet of the first tank
comes from Qa(which is SNO in the fifth reactor (SNO,5)), Qa
increases cause SNO in the input of the first reactor (SNO,0)
increases (14). The SNH,0 and SNO,0 values affect SNH,e and
SNtot,e values after a period of time that depends on HRT and
therefore on the flow rate.
SNH,0=Qin ·SNH,po +Qa·SNH,5
Qin +Q(13)
SNO,0=Qin ·SNO,po +Qa·SNO,5
Qin +Qa
(14)
2) Qaeffect in the denitrification process: In the
denitrification process that takes place in anoxic reactors, SNO
is reduced to molecular nitrogen, which is harmless. XB,H
consume SSusing oxygen from SNO (due to the absence
of SO). Thus, as it can be seen in the rSNO equation (2, 5),
the greater the amount of SS, the greater the SNO reduction.
Due to the fact that SSis reduced during the biological
treatment, an increase in Qareduces SNO in the anoxic
reactors, worsening denitrification.
A more important effect of Qaon the denitrification
process is its influence on HRT, since when the flow
3
590
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
is higher, HRT decreases and vice versa. As can been
observed in the mass balance equation (11), an increase in
Qadiminishes the effect of rZand therefore worsens the
denitrification process, increasing SNO and consquently SNtot,e.
3) Qaeffect in the nitrification process: The nitrification
process takes place in aerobic reactors, where XB,A oxidize
SNH into SNO. In the same way as in the denitrification
process, Qaincreases attenuate the effect of rZ(11),
worsening in this case the nitrification process, which causes
SNH increase and SNO decrease. However, SOis often
regulated in the aerobic reactors based on SNH. In this case,
Qaincreases can result in an SOincrease to improve the
nitrification process, which reduces SNH and increases SNO.In
conclusion, the best way to reduce SNH is to reduce Qa.
Regarding SNtot, as its main components are SNH and SNO,a
trade-off solution must be found between both to reduce SNtot.
B. Fuzzy controller design
Fuzzy logic can be defined as a control based on human
expertise. Fuzzy controller adapts the input and output
variables into suitable linguistic values by membership
functions. Rules between input and output variables are
established by words. Readers can found further information
about fuzzy control in standard references as [21]. The FIS1
Editor from Matlab is used for the implementation of the
proposed fuzzy controller.
As explained in Section III-A3, to assess the Qaeffects on
the biological treatment requires an exhaustive knowledge of
the plant behavior. Due to this reason, a fuzzy controller has
been proposed for the Qamanipulation.
The proposed fuzzy controller has 6 inputs, 1 output and
30 rules. As shown in Fig. 5, the inputs are SNH in the
influent (SNH,in), SNH,0 ,SNH,5,SNO,5 ,T
as and influent flow
rate (Qin) and the output is Qa. Mamdani ( [22]) is the
method of inference. The rules are based on the effects of
Qaon the biological treatment explained in Section III-A3,
and are detailed below. Some of them are also shown in the
surface graphs (Fig. 2 and 3).
In dry weather, Qin values are related to SNH,in. When
this relationship does not exist, due to an increase in Qin
with respect to SNH,in, a rain event is considered. During
dry weather, when SNH,in is ”low”, Qais decreased to
reduce pumping energy costs and improve nitrification and
denitrification processes., and if SNH,in increases, Qais also
increased in order to dilute SNH,0. In the case that a rain
event is detected, Qais increased to dilute SNH,0, as long as
there is no risk of SNH,5 or SNO,5 increase. Fig. 2a shows the
relationship between Qin and SNH,in.
1FIS : Fuzzy Inference System
Both during dry and rainy weather (except when SNH,in is
”low” in dry weather), Qavalues are inversely based on SNO,5
values, such that the higher the SNO,5, lower is Qa(Fig. 2b).
This is because the proposed fuzzy controller can be applied
with control strategies that regulate SOset-point based on
SNH, since better results are obtained instead of keeping SO
set-point fixed ( [10]) . So, decreasing Qa,SNH,5 is reduced
and consequently also SOand SNO,0. It should be noted that
this fact does not happen in the case that SNH,5 is so high
that the value of SO,5 reaches its maximum value defined by
the controller that regulates it, even if Qais reduced. In the
case of control strategies with fixed SOset-point, the fuzzy
controller design should be modified by adding SOas input.
Qavalues depend on Tas, with both dry and rainy weather,
since affects the denitrification and nitrification processes.
When Tas decreases, these processes worsen, as shown in
the equations of observed conversion rates (2, 1, 5, 6, 4, 7,).
Therefore, Qais higher when Tas is lower with the aim of
diluting more SNH,0 (Fig. 2b).
Qareduction is always limited by SNH,0. When it is not
necessary to dilute SNH,Qais reduced to reduce pumping
costs and to improve nitrification and denitrification processes.
The SNH,0 value limits the Qareduction, since there must be
a minimum SNH dilution in order to make possible to reduce
it in the nitrification process to levels below the established
limits (Fig. 3c).
Finally, SNH,5 always plays a priority role in the rules, so
that the Qaincrease is always limited by it, because when
SNH,5 is ”high”, Qais ”low” or ”very low” (depending on
Tas) (Fig. 3d).
C. Control strategy where the proposed fuzzy controller is
tested
The proposed fuzzy controller is applied in the default
Control Strategy (DCS) with SOhierarchical control.
The original BSM2 definition ( [18]) proposes a DCS.The
closed-loop control configuration consists of a PI that controls
the SO,4 at a set-point of 2 mg/l by manipulating oxygen
transfer coefficient (KLa) in the third tank (KLa3), KLain
the fourth tank (KLa4) and KLain the fifth tank (KLa5)
with KLa5set to the half value of KLa3and KLa4. Carbon
addition (qEC) in the first reactor (qEC,1 ) is added at a constant
flow rate of 2 m3/d. Two different wastage flow rate (Qw)
values are imposed depending on time of the year: from 0 to
180 days and from 364 to 454 days Qwis set to 300m3/d;
and for the remaining time periods Qwis set to 450m3/d.
Finally, Qais kept fixed to 61944 m3/d (Fig. 4).
As explained in Section (III-B), the proposed fuzzy
controller is designed for control strategies with regulated
SOset-point. For this reason, a hierarchical control for SO,4
4
591
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
(a) X input: Qin; Y input: SNH,in ; Z Output: Qawith SNH,0
”low”, SNH,5 ”not high”, SNO,5 ”medium” and Tas ”low” and
”high” (15C)
(b) X input: SNO,5; Y input: Tas ; Z Output: Qawith SNH,in
”medium”, SNH,0 ”low”, SNH,5 ”not high” and Qin ”low”
Fig. 2. Graphic surfaces of the fuzzy control output related to the inputs
set-point manipulation is added to DCS by means of a fuzzy
controller designed and tested in ( [10]) for BSM1 and
adapted to the BSM2 characteristics in ( [14]) . This referred
paper shows that better results are obtained by regulating SO
based on SNH, instead of trying to keep SOat a fixed set-point.
For testing the proposed fuzzy controller, it has been added
to this control strategy, replacing it in place of the fixed Qa
value (Fig. 5).
IV. SIMULATION RESULTS AND DISCUSSION
Table II shows the results obtained with DCS with SO
hierarchical control, both with fixed Qa, and manipulating
Qawith the proposed fuzzy controller, so the improvement
obtained by the proposed fuzzy controller can be observed.
Regarding the effluent quality, the greatest improvement is
obtained in the reduction of SNH,e limit violations, by 68.78%.
SNtot,e limit violations are also reduced by 24.92%. The limit
violations of COD, TSS and BOD5remain the same, as is
logical given that the proposed fuzzy controller is designed
only to reduce limit violations of SNH,e and SNtot,e. It should
be mentioned that, COD, TSS and BOD5limit violations
happen when the bypass is active, due to a significant Qin
(a) X input: Qin; Y input: SNH,0 ; Z Output: Qawith SNH,in
”medium”, SNH,5 ”not high”, SNO,5 ”medium” and Tas ”low”
and ”high” (15C)
(b) X input: SNH,5 ; Y input: SNH,in; Z Output: Qawith SNH,0
”low”, SNO,5 ”medium”, Qin ”low” and Tas ”low” and ”high”
(15C)
Fig. 3. Graphic surfaces of the fuzzy control output related to the inputs
!!
!!!!!!!! !!!!!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
! ! ! ! ! ! ! ! ! !
!
!
!
!!!!!!!! ! ! !!!!!!!
!
!!!!!!
!
!!!!!!!!!!!!!!!!!!
Qa!=!61944!m3!/!day!!
6161\begin{figure}[h
!
WWTP!
!
PI!
SO,4!set-point!
KLa3!
KLa4!
1/2!
KLa5!
SNH,5!!
Fuzzy!for!SO!hierarchical!
SO,4!!
Fig. 4. DCS with SOhierarchical
increase that the plant cannot assume. Moreover, the reduction
of limit violations does not imply an increase in operational
costs, since there is an OCI reduction of 0.19%, mainly due
to a pumping energy saving of 3.79%. The reasons for these
improvements can be observed in figures 6, 7, 8 and 9 and
are discussed below.
Fig. 6 shows the time evolution for the main concentrations
during day 341. It is observed how SNtot,e is reduced with the
proposed fuzzy controller compared to Qafixed and it is kept
below the established limit. The SNtot,e increase is due to a
previous SNH,in increase and a subsequent Qin increment that
makes nitrification and denitrification processes worse. When
5
592
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
!!
!! ! ! ! ! ! ! ! !!!!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
! ! ! ! ! ! ! ! ! !
!
!
!
! ! ! !!!!! ! ! !!!!!!!
!
!!!!!!
!
! ! ! ! !!!!!!!!!!!!!!
Qa!
!
WWTP!
!
PI!
SO,4!set-point!
KLa3!
KLa4!
1/2!
KLa5!
SNH,5!!
Fuzzy!for!SO!hierarchical!
Proposed(fuzzy(controller(
for(Qa(manipulation((
Qin!!
Tas!!
SNO,5!!
SNH,0!!
SNH,in!!
SO,4!!
Fig. 5. DCS with SOhierarchical and the proposed fuzzy controller for Qa
manipulation
there is a SNH,in increase, Qais increased with the aim of
diluting SNH,0. When SNO,5 increases, Qais slightly reduced,
improving the nitrification process, being less SOrequired
and consequently generating less SNO,5. The previous dilution
of SNH,0 and the subsequent reduction of Qaallow SNtot,e to
be kept below the required limit. The reduction in Qais slight
due to the high levels of Qin. In fact, Qais subsequently
increased because Qin reaches usual levels of rain events. It
should be note that an excessive Qareduction can increase
SNO,5 by improving the nitrification process.
Fig. 7 corresponds to day 368. On the one hand a similar
situation to that commented in the previous paragraph on Fig.
6 is observed, to reduce the Ntot,e peak below the established
limit. In this case, the Qaincrease is earlier, since there is
aQin increase with respect to SNH,in, thus considering that
there is a rain event. On the other hand, the Qareduction that
occurs later is greater than in Fig. 6, because it is produced
by a higher increase in SNH,5 and aims to avoid a SNH,5 limit
violation, improving the nitrification process. Therefore, in this
case the reduction occurs before SNO,5 increases. As shown
in Fig. 7, the proposed fuzzy controller avoids SNH,5 limit
violation, while keeping Qafixed there is SNH,5 limit violation.
The costs reduction is mainly due to the reduction in
pumping energy due to lower average Qa. As seen in Fig. 8
it occurs mainly in summer considering that Tas is ”high”,
beacuse nitrification and denitrification processes improve
and the necessary SNH,0 dilution is less than in winter to keep
both SNtot,e and SNH,e below the established limits. Fig. 8
shows how most of the time Qais lower using the proposed
fuzzy controller than with fixed Qa. By manipulating Qawith
the proposed fuzzy controller, SNtot,e is usually higher than
with fixed Qa, but much lower than the established limit. The
regulation of Qahas the objective of reducing the peaks of
SNH,e and SNtot,e although the dilution of SNH,0 is lower.
Fig. 9 shows a week when Tas is ”low”. By worsening
the nitrification and denitrification processes, when SNH,in
increases the proposed fuzzy controller increases more
Qathan when Tas is ”high” in order to dilute SNH,0.
The SNtot,e and SNH,e peaks are higher than in summer
and are closer to the established limits. Using the proposed
fuzzy controller both peaks are reduced compared to fixed Qa.
Fig. 6. Time evolution of Qin,SNH,in ,SNH,0,SNO,5 ,SNH,5,SNtot,e and Qa
of day 341 for DCS with SOhierarchical with fixed Qaand manipulating Qa
with the proposed fuzzy controller.
Fig. 7. Time evolution of Qin,SNH,in ,SNH,0,SNO,5 ,SNH,5,SNtot,e and Qa
of day 368 for DCS with SOhierarchical with fixed Qaand manipulating Qa
with the proposed fuzzy controller.
V. C ONCLUSIONS
This article has presented a new control strategy to manip-
ulate Qataking into account its effects in the different areas
6
593
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
TABLE II
SIMULATION RESULTS OF THE DEFAULT CONTROL STRATEGY WITH SOHIERARCHICAL CONTROL AND THE SAME CONTROL STRATEGY ADDING THE
PROPOSED FUZZY CONTROLLER FOR QAMANIPULATION
Evaluation Criteria DCS+SOhierarchical control DCS+SOhierarchical control with % of
the proposed fuzzy controller for Qamanipulation improvement
Effluent Quality SNtot,e limits violations (% of time) 0.309 0.232 24.919
SNH,e limits violations (% of time) 0.229 0.0715 68.777
COD limits violations (% of time) 0.0572 0.0572 0
TSS limits violations (% of time) 0.343 0.343 0
BOD5limits violations (% of time) 0.226 0.226 0
Operational Costs Pumping energy (kWh/day) 445.454 428.558 3.793
Sludge production (kg SS/day) 2707.477 2708.805 -0.049
Areation energy (kWh/day) 3601.86 3596.133 0.159
Carbon source dosage (kg COD/day) 2400 2400 0
Mixing energy (kWh/day) 769.113 768.484 0.082
Heating energy (kWh/day) 0 0 0
Net energy production from methane (kg CH4/day) 1086.239 1085.865 0.034
OCI 8821.425 8.804.404 0.193
Fig. 8. Time evolution of Qin,SNH,in ,SNH,0,SNO,5 ,SNH,5,SNtot,e and Qa
during one week in summer for DCS with SOhierarchical with fixed Qaand
manipulating Qawith the proposed fuzzy controller.
of the biological treatment of a wastewater treatment plant, by
applying a fuzzy controller. The proposed controller has been
tested in DCS with hierarchical SO, replacing it and comparing
it with keeping Qafixed (as in DCS). The graphs of evolution
over time of the most important concentrations show the
effects of Qaat the different concentrations, corroborating the
objectives of the proposed fuzzy controller. In general terms,
increasing Qato dilute SNH,0 when SNH,in or Qin increases and
its reduction when SNH,5 or SNO,5 increase, allow to reduce
limits violations of SNH,e and SNtot,e compared to keeping
Qafixed. On the other hand, the Qadecrease when it is
not necessary to dilute SNH,0 due to low SNH,in levels and
especially in summer, reduces the energy costs of pumping
compared to keeping Qafixed. Final results have shown the
reduction of SNH,e and SNtot,e limit violations and operational
costs. Future work will be aimed at testing the proposed fuzzy
controller in other already published control strategies.
Fig. 9. Time evolution of Qin,SNH,in ,SNH,0,SNO,5 ,SNH,5,SNtot,e and Qa
during one week in winter for DCS with SOhierarchical with fixed Qaand
manipulating Qawith the proposed fuzzy controller.
ACKNOWLEDGMENTS
This work was partially supported by the Spanish CICYT
program under grant DPI-2016-77271-R.
REFERENCES
[1] M. Barbu, I. Santin, and R. Vilanova, “Applying control actions for water
line and sludge line to increase wastewater treatment plant performance,”
Industrial & Engineering Chemistry Research, vol. 57, no. 16, pp. 5630–
5638, 2018.
[2] I. Sant´
ın, R. Vilanova, M. Barbu, and C. Pedret, “Manipulating the
sludge flow to increase methane production and to reduce effluent
pollution in wastewater treatment plants,” in 2016 24th Mediterranean
Conference on Control and Automation (MED). IEEE, 2016, pp. 877–
882.
[3] C. A. C. Belchior, R. A. M. Ara´
ujo, and J. A. C. Landeck, “Dissolved
oxygen control of the activated sludge wastewater treatment process us-
ing stable adaptive fuzzy control,” Computers & Chemical Engineering,
vol. 37, pp. 152–162, 2012.
[4] T. Yang, W. Qiu, Y. Ma, M. Chadli, and L. Zhang, “Fuzzy model-based
predictive control of dissolved oxygen in activated sludge processes,”
Neurocomputing, vol. 136, pp. 88–95, 2014.
7
594
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
[5] G. Harja, I. Nascu, C. Muresan, and I. Nascu, “Improvements in
dissolved oxygen control of an activated sludge wastewater treatment
process,” Circuits, Systems, and Signal Processing, vol. 35, no. 6, pp.
2259–2281, 2016.
[6] X. Du, J. Wang, V. Jegatheesan, and G. Shi, “Dissolved oxygen control
in activated sludge process using a neural network-based adaptive pid
algorithm,” Applied sciences, vol. 8, no. 2, p. 261, 2018.
[7] I. Sant´
ın, M. Barbu, C. Pedret, and R. Vilanova, “Dissolved oxygen
control in biological wastewater treatments with non-ideal sensors
and actuators,” Industrial & Engineering Chemistry Research, vol. 58,
no. 45, pp. 20 639–20 654, 2019.
[8] P. Vega, S. Revollar, M. Francisco, and J. Mart´
ın, “Integration of set
point optimization techniques into nonlinear mpc for improving the
operation of wwtps,” Computers & Chemical Engineering, vol. 68, pp.
78–95, 2014.
[9] K. Błaszkiewicz, R. Piotrowski, and K. Duzinkiewicz, “A model-based
improved control of dissolved oxygen concentration in sequencing
wastewater batch reactor,” Studies in Informatics and Control, vol. 23,
no. 4, pp. 323–332, 2014.
[10] I. Sant´
ın, C. Pedret, and R. Vilanova, “Applying variable dissolved
oxygen set point in a two level hierarchical control structure to a
wastewater treatment process,” Journal of Process Control, vol. 28, pp.
40–55, 2015.
[11] F. Li, J. Qiao, H. Han, and C. Yang, “A self-organizing cascade neural
network with random weights for nonlinear system modeling,” Applied
soft computing, vol. 42, pp. 184–193, 2016.
[12] S. Revollar, P. Vega, R. Vilanova, and M. Francisco, “Optimal control of
wastewater treatment plants using economic-oriented model predictive
dynamic strategies,” Applied Sciences, vol. 7, no. 8, p. 813, 2017.
[13] J.-F. Qiao, Y. Hou, L. Zhang, and H.-G. Han, “Adaptive fuzzy neural
network control of wastewater treatment process with multiobjective
operation,” Neurocomputing, vol. 275, pp. 383–393, 2018.
[14] I. Sant´
ın, C. Pedret, R. Vilanova, and M.Meneses, “Advanced decision
control system for effluent violations removal in wastewater treatment
plants,” Control Engineering Practice, vol. 279, pp. 207–219, 2016.
[15] K. Gernaey, U. Jeppsson, P. Vanrolleghem, and J. Copp, “Benchmarking
of Control Strategies for Wastewater Treatment Plants,” IWA Publish-
ing,, London, UK, Scientific and Technical Report No.23,, 2014.
[16] J. Alex, J. Beteau, J. Copp, C. Hellinga, U. Jeppsson, S. Marsili-Libelli,
M. Pons, H. Spanjers, and H. Vanhooren, “Benchmark for evaluating
control strategies in wastewater treatment plants,” in Control Conference
(ECC), 1999 European. IEEE, 1999, pp. 3746–3751.
[17] J. B. Copp, The COST Simulation Benchmark: Description and Simu-
lator Manual: a Product of COST Action 624 and COST Action 682.
EUR-OP, 2002.
[18] U. Jeppsson, M.-N. Pons, I. Nopens, J. Alex, J. Copp, K. Gernaey,
C. Rosen, J.-P. Steyer, and P. Vanrolleghem, “Benchmark Simulation
Model No 2: general protocol and exploratory case studies,” Water
Science and Technology, vol. 56 (8), pp. 67–78, 2007.
[19] I. Nopens, L. Benedetti, U. Jeppsson, M.-N. Pons, J. Alex, J. B.
Copp, K. V. Gernaey, C. Rosen, J.-P. Steyer, and P. A. Vanrolleghem,
“Benchmark Simulation Model No 2: finalisation of plant layout and
default control strategy,” Water Science and Technology, vol. 62 (9), pp.
1967–1974, 2010.
[20] M. Henze, C. Grady, W. Gujer, G. Marais, and T. Matsuo, “Activated
sludge model 1,” IAWQ,, London, UK, Scientific and Technical Report
No.1,, 1987.
[21] G. Klir and B. Yuan, Fuzzy sets and fuzzy logic. Prentice hall New
Jersey, 1995, vol. 4.
[22] E. Mamdani, “Application of Fuzzy Algorithms for Control of Simple
Dynamic Plant,” Proceedings of the Institution of Electrical Engineers,
vol. 121 (12), pp. 1585–1588, 1976.
8
595
Authorized licensed use limited to: Universitat Autonoma De Barcelona. Downloaded on April 22,2021 at 08:37:38 UTC from IEEE Xplore. Restrictions apply.
