CONTROL STRATEGIES FOR ACTIVATED SLUDGE TREATMENT SYSTEM
By
Bhakta Kabi Das
and
Pratap K. J. Mohapatra
Abstract
The activated sludge treatment system for treating municipal wastewater presents an interesting
application of system dynamics modeling. This paper presents such a modeling approach to the
strategy formulation of the treatment system in order to economically control effluent quality.
First, factorial designs are carried out on the simulation results to identify factors that
significantly affect effluent quality. Thereafter, open-loop control (both constant and time-
varying), output feedback control, and output-integrated feedback control strategies have been
applied. Statistical tests of significance indicate that the strategy of output feedback control has
the maximum potential, in both summer and winter, to achieve the dual objectives of maintaining
effluent quality within acceptable limits and minimizing aerator energy.
Principles Underlying the Activated Sludge Treatment System
The process of Activated Sludge Treatment System (Fig. 1) is the most widely employed
technique today it for treatment of municipal wastewater. The influent stream of municipal
wastewater, rich in soluble organic compounds (known as substrate or food for the bacteria),
enters the reactor (known as aeration tank). Bacteria feed on the organic waste present in the
water. Aeration is done by mechanical means providing the much needed oxygen required for the
growth of bacteria in the tank.
Secondary Effluent
Clarifier
Influent Aeration ee
tank
>
Active Recirculated Excess Sludge
Sludge
Figure 1: Schematic Diagram of Activated Sludge Treatment System.
The sewage stream loaded with biomass goes out of the aeration tank and reaches the clarifier.
The clarifier allows the biomass to flocculate and settle down, due to gravity, as sludge. The
sludge contains live biomass. It is re-circulated, in a controlled manner, to the aeration tank so as
to maintain the biomass concentration in the tank at a desired level. The excess sludge, which is
not re-circulated, is withdrawn from the clarifier and put to sludge drying beds for future use as
manures. The clarified supernatant stream goes out of the system as treated effluent.
Researchers have developed a number of models for studying and understanding the wastewater
treatment process. These models can be broadly categorized as (1) Component level models and
(2) Comprehensive models. Component-level models pertain to sub-areas such as ‘substrate
removal process’ (Novak 1974, Lawrence and McCarty 1974), ‘biomass growth process’ (Grady
and Roper, Jr. 1974, Gaudy, Jr., et al. 1974, Grady et al. 1986) and Monod 1949), ‘oxygen
transfer, dissolved oxygen consumption, and oxygen assimilation process’ (Bliss and Barnas
1986, Picionreanu, et al. 1997), and ‘clarification process’ (Ford and Eckenfelder 1967).
Comprehensive models pertain to the functioning of the whole treatment system and can be
classified as analytical models (Ford and Eckenfelder 1967, Roper, Jr. and Grady, Jr. 1978,
Smeers and Tyteca 1984, Uber, etal. 1985, Tang, etal. 1987, Zhao, etal. 1999, and Anderson, et
al. 2000), simulation models (Busby and Andrews 1975, Barton and Mckeown 1986, and
Anderson, et al. (2000)), and system dynamics models (Das, etal. 1995, 1997 and Clemson, et al.
1995).
But for Anderson, et al. (2000), none of the studies reported above has designed any control
strategy for the output quality exceedences. Clemson, et al. (1995) have developed a system
dynamics model for wastewater treatment plant and have used Taguchi methods in conducting
sensitivity analysis. But their model considers neither the biological process of growth of
microorganism nor the process of oxygenation. It also does not try to design a strategy for
effluent quality control.
The present work presents a comprehensive system dynamics model for wastewater treatment
plant by using the tools, techniques and concepts of design of experiments, statistical quality
control, and modern control theory in order to decide the number of aerators to use and design the
sludge re-circulation policy while maintaining the effluent quality within acceptable limits.
A Dynamic Model for the Treatment System
Four physical flows can be distinguished in an activated sludge plant: (1) Flow of Liquid, (2)
Flow of Biomass, (3) Flow of Substrate or Pollutants, and (4) Flow of Dissolved Oxygen.
Flow of Liquid
Figure 2 is the causal-loop diagram for the flow of liquid. It shows that the inflow to the aeration
tank increases the liquid accumulation in the tank and causes an increase in the outflow from the
tank (since the tank is always full). This, in turn, decreases the liquid accumulation in the tank.
The treated outflow from the tank increases the sludge quantity settled in the clarifier and
subsequently increases the clarified effluent quantity, waste sludge quantity, and re-circulated
active sludge quantity.
Sewage } » -
Inflow . O 2
/ {o) 4
;
; 2) ~~ Tmatedt
Aciive sucge me se
reareu sien putt ow trorr
aeration tank
Treated ae
ywage in
Bi A Chanter
(3) _- effluent
Figure 2 Causal Loop Diagram for Sewage Quantity
Figure 3 is the causal loop diagram for the flow of Biomass. With the inflow of wastewater, the
biomass inflow increases, resulting in a rise in total biomass in the aeration tank. This rise in the
level of biomass raises the biomass generation rate within the tank, resulting in a positive
feedback loop. The increase of biomass leaving rate through the outflow from the tank causes an
increase of biomass in the clarifier and an increase of biomass leaving rate through the clarified
effluent. The increase of biomass trapped in the clarifier increases the formation of sludge and
increases both the biomass re-circulated as active sludge and the biomass leaving the system
through the waste sludge. However, both the outflows reduce the biomass trapped in the clarifier.
Biomass,
in aeration
tank
O
4)
Biomass (*)
through
Biomass
influent
; leaving AT.
Biomass
Biaiiaes generation
recirculated
as active i]
surge
Sewage
inflow
) Pan. Biomass
Biomass in Leaving
through
clarifier Clari
¥ arified
Biomass Q
leaving the Biomass’ effluent
system ‘~~ trapped in
through 0 clarifier to
waste form
sludge (0) sludge
Figure 3 Causal Loop Diagram of Biomass in Activated Sludge Treatment System
F low of Substrate
Figure 4 is the causal-loop diagram for the flow of substrate. It shows that as the inflow of
sewage increases, the substrate coming with the influent increases and so does the substrate
accumulation in the aeration tank that enhances the specific biomass growth rate and substrate
outflow through the effluent. A rise in the specific biomass growth rate raises the specific
substrate consumption rate and lowers the substrate accumulation level in the aeration tank.
F low of Dissolved Oxygen
Figure 5 is the causal-loop diagram for the dissolved oxygen (DO) in the activated sludge
treatment system. The atmospheric oxygen gets transferred to the wastewater environment with
the help of surface aerators. Therefore any rise in oxygen transfer rate raises the DO
concentration in the aeration tank. This outflow from the aeration tank depletes the DO level in
the aeration tank, thereby reducing the DO concentration at the aeration tank.
Substrate
through
influent Substrate
accumulation
in aeration.
tank
Substrate
Consumption
Sewage rate
inflow
0)
Substrate
cone in AT.
Substrate
Outflow
Specific through
Substrate Specific effluent
Consumption Biomass
rate a ee Growth
Figure 4 Causal Loop Diagram of Substrate in Activated Sludge Treatment System
The reduction in DO concentration at the aeration tank increases the gap between the oxygen
level in the water environment and that in the air environment. This increases the oxygen
assimilation rate and the oxygen transfer rate. The DO level in the aeration tank decreases as the
biomass utilizes it for respiration.
The well-known equation of Monod (Monod 1949) has been used here to model biomass growth
rate and past works by Arceivala (1981), Grady And Lim (1980), and by Sincero and Sincero
(1996) are used to model the dissolved oxygen transferring ability of the aerators, oxygen
assimilation capacity of waste water, and the dissolved oxygen utilization rate by the biomass.
The Base Run
The following considerations are made for the base model run.
1. The biomass culture is a unique mixture, viable and typically acclimatized with domestic
sewage in aerobic environment.
2. The pollution (substrate) is of soluble and readily biodegradable carbonaceous nature,
and is measured in terms of Chemical Oxygen Demand (COD) expressed as kg of
COD/m’.
3. The extended aeration version of activated sludge treatment system is considered for
modeling, with an average hydraulic retention time of 16 hours.
yN OS
Oxygen DO conc at DO outflow
Transfer rate aeration tank frorn aeration
tank
Oxygen
assimilation DO level at
rate aeration tank
DO utilization
rate
Figure5: | Causal loop diagram of Dissolved Oxygen in Activated Sludge Treatment
System
4. The volume of sewage in aeration tank is 3200 m’ and is full with sewage at the start.
5. Initial average inflow rate of sewage is 200 m*/hr.
6. The initial influent biomass concentration is 0.002 kg of COD/m’ and the initial influent
substrate concentration is 0.15kg of COD/m’.
7. Following Mynhier and Grady (1975), the typical saturation substrate constant and the
typical maximum specific biomass growth rate are taken as 0.06 kg of COD/m’ and 0.13
kg/kg/hr respectively; the typical biomass decay coefficient is taken as 0.003 per hour;
and the typical biomass decay coefficient is taken as 0.003 per hour.
8. The dissolved oxygen and other nutrients such as Nitrogen and Phosphorous are assumed
to be maintained continuously at a level adequate for proper biomass growth.
9. The sludge recirculation ratio is kept constant at 0.8.
The Powersim software package was used to simulate the model. The model was run for 200
hours. Extensive validation testing was done and the results obtained were plausible. Following
initial results that appeared to be favourable, the sludge recirculation ratio was taken constant at
0.65. For the purpose of model validation, a modular rise in the value of aerators-in-use was
considered: 0, 50 hp, 100 hp, and 150 hp.
Figure 6 shows the graphical response and a portion of simulated values in a tabular format when
‘Aerators_in_use’ is zero and a uniform average inflow with influent substrate concentration of
0.15 kg/ m? keeps on coming during the test period. The result shows that after an initial
transience the effluent quality equals the influent quality. In the absence of aerators oxygen is in
short supply and this causes drastic reduction in biomass growth and hence the effluent quality
does not improve in this case.
Aerators_in_use = 0 hp
Aerators_in_use =O0hp Influent Quality = 0.15 kg/cum
Influent Quality = 0.15 kg/cum Time Effluent_Quality
0.20-
°
a
°
6
Effluent Quality[kg/cum]
°
R
°
=I
S
oF
50 100 150 200
Timefhr]
Figure 6 Effluent Quality response for Aerators_in_use = Ohp
Figure 7 shows the graphical response and a portion of simulated values in a tabular format for
the run results considering the capacity of ‘Aerators_in_use’ as 50 hp. The system stabilizes after
almost fifty hours and the effluent quality improves as expected and achieves an average value of
0.032 kg/m, about five-time improvement over the influent quality.
Table 1 shows the average final values of effluent quality for different values of aerators in use.
We see that in general when aerators_in_use increases, the effluent quality improves. However,
for 150 hp aerators in use, the effluent quality deteriorates. A reason for this to happen is that
higher availability of oxygen raises the biomass level to such a high level that food is not
sufficient to sustain their growth. This results in a fall in the biomass level and deteriorates the
effluent quality.
Aerators_in_use= 50 hp
Aerators_in_use= 50 hp Influent Quality = 0.15 kg/cum
Influent Quality = 0.15 kg/cum Tine Effluent_Guality
= 0.20. 1nn nna?
2 101 0.0292
® 04s 102 0.0265
= 103 0.0245
S 0410 104 0.0235
a 105 0.0234
5 ons wre et 106 0.024
e SVN 107 0.0255
ooo | 108 _| 0.0276
0 50 100 150 200 109 0.03
Timefhr] 110 0.0325
Figure 7 Effluent Quality response for Aerators_in_use = 50 hp
Table 1; Final Effluent Quality for Different Aerators_in_Use
Aerators_in_use (hp) Avg Final Effluent Quality (kg/m*)
0 0.15
50 0.032
100 0.029
150 0.0298
The test results indicate also that for the given set of influent conditions, the maximum total
capacity of aerators should be limited to 100 hp in order to achieve a satisfactory effluent quality.
Analysis of Simulation Experimental Results: The 3” Design
In order to find out the extent of influence the aerators in use and the recirculation ratio have on
the effluent quality, we now simultaneously change the values of the aerators in use and the
recirculation ratio. For the purpose of analysis of simulation results, we consider these two
factors, and consider three levels for each of these factors:
Factors Levels
Aeration-in-use 50 hp, 100 hp, 150 hp
Sludge Recirculation Ratio 0.1, 0.5, 0.8
For each combination of factors, we replicated the simulation experiment four times with
different noise levels. To induce the desired noises, (1) normal random noises were introduced in
four variables: (a) Influent substrate concentration, (b) Influent dissolved oxygen concentration,
(c) Influent biomass concentration, and (d) Net biomass production rate, and (2) different seeds
were used in the random number generators.
Noises were assumed to follow normal distribution with zero mean. The standard deviations
were chosen in a manner such that the steady state values of the uncontaminated variables were at
least 4 times the standard deviations of the noises, so as not to result in any unrealistic negative
values of the variables in the presence of the noise.
The model was simulated under the conditions stated above. Analysis of variance was done on
the simulation results. Table 2 gives the ANOVA table. It is seen that the effect of aeration-in-
use, sludge recirculation ratio, and their interaction effect are significant in explaining the
variation in the average effluent quality. From the F-values it is inferred that the recirculation
ratio has the highest leverage in explaining the variations in effluent quality, followed by the
aerators in use and their interaction.
Table 2: ANOVA Table for 3” Design
Source of Variation Sum of df | Mean Sq. F- _MS_ Remarks
Squares —- ss ° VB enor
ss df
SL. Recirculation Ratio | 0.0030265 2 | 0.0015132 408564
Aerators-in Use 0.0004521 2 | 0.000226 61020 Serror =
Interaction 0.0007343 4 | 0.0001835 49545 0.0000608
Error 0.0000001 | 27 | 3.7037x10
Total 0.0042131 | 35
Fo 05, 4,27 = 2.73 ; Foos,2,27 = 3.35
CONTROL STATEGIES
The operational strategies are to be developed for (1) controlling the exceedence of the effluent
quality from the permissible output quality and (2) minimizing the aerator energy consumption.
Following the commonly recommended control strategies in modern control theory two broad
categories of strategy are adopted here: (1) Open-loop control and (2) Closed-loop control.
Open-loop Control
In the open-loop control strategy, the feedback information on the effluent quality is not used to
design the control variables. We have used the following two types of open-loop control
strategies: (1) Constant control and (2) Time-varying control.
Constant Control
Here the two control variables (aerators-in-use and sludge recirculation ratio) were held constant
during the entire simulation run of the model. The following values were assumed for the two
control variables:
Aerators-in-Use : 50 hp, 75 hp, and 100 hp.
Sludge Recirculation Ratio : 0.1, 0.5, 0.65, and 0.8.
Following the temperature variations normally experienced in the eastern coastal India, the model
assumed the temperature to vary from a minimum of 8°C at 12 midnight to a maximum of 16°C at
noon during winter and from a minimum of 18°C at midnight to a maximum of 24°C at noon
during summer. Mimicking the normal pattern of diurnal variations of inflow rate in municipality
water systems, the model assumed the inflow rate to vary during a day, taking a minimum of 75
cubic meters per hour at 00 hour and a maximum of 250 cubic meters per hour at 08 hour. The
model was tested with various factor combinations of diurnal variations in temperature and
inflow rate.
The simulation results for the effluent quality for the period 100-200 hours (neglecting the initial
transient period between 0-100 hours) were transferred to an Excel file through the DDE
[Dynamic Data Exchange] facility of the Powersim Package to facilitate statistical computation.
The criteria for selecting the best policy are given below in the decreasing order of priority:
¢ Meeting the effluent quality norm, i.e., minimizing the deviation of the mean steady-state
effluent quality from its target value.
e Minimizing the standard deviation of the effluent quality.
e Minimizing the energy expended in the aerators.
Figure 8 shows, for summer, the variations in the mean effluent quality with the variation in
recirculation ratio for different values of aerators-in-use. Figure 9 shows similar results for
winter. As sludge recirculation rate rises, bacteria level rises in the aeration tank. It degrades the
sludge and improves the quality. Beyond a recirculation ratio value of 0.65, however, the biomass
population becomes excessive, going beyond the biomass growth sustainable by the amount food
in the wastewater, and thus deteriorates the effluent quality.
To find out the best policy for each season, statistical tests of hypothesis are carried out (reported
elsewhere, Das 2003). The tests indicate that the strategy of using a sludge recirculation ratio of
0.65 and aerators-in-use of 100 hp gives the best effluent quality for each season.
Time-Varying Control
Table 3 gives the configuration of aerator-in-use under time-varying control strategy. Taking a
cue from the constant-control results, the sludge recirculation ratio was kept at a constant value of
0.65. The capacities of aerators to be deployed at various times of the day are selected to take care
of the diurnal variations of temperature and inflow rate. Thus, the highest capacity of aerators is
deployed during the 08-20 hours and the lowest capacity was deployed during 00-04 hours.
Considering that testing of effluent quality for COD values involves digestion, cooling and
titration and requires about 4 hours, the inflow rate variations were considered in a time slot of 4
hours in a day. The configuration of aerators in use for the time-varying control strategy
followed for the model simulation is shown in Table 3.
The model results for the effluent quality are taken to an Excel file and the corresponding
statistical values, mean and standard deviation are calculated. The total aerator energy (hp-hr)
consumed during the period 100-200 hour is given in Table 4. A study of the results reported in
Table 4 indicates that all policies 1, 3, and 5 for winter give acceptable effluent quality values, but
the aerator energy is the minimum for policy 1. Thus policy | is considered the best for winter.
For summer, however, only policy 6 (out of the policies 2, 4, and 6) gives effluent quality values
within acceptable range. Thus policy 6 is considered the best policy for summer, although it
requires the maximum aerator energy.
2
So
a
0.055 | 2-50 hp
Mean Effluent Quality (kgim’*)
0.03 :
0.1 03 05 0.65 0.7 og O09
Recirculation Ratio (dimensionless)
Figure 8 Mean Effluent Quality in Summer
0.05
0.045
0.04
0.035
Mean Effluent Quality (kgim’)
0.03
0.1 0.3 05 065 07 O8 O09
Recirculation Ratio (dimensionless)
Figure 9 Mean Effluent Quality in Summer
Table 3: The Configuration of Aerator-in-use under Time-Varying C ontrol Strategy
Policy Sludge Season Aerator-in-Use (hp)
No. Recirculation Ratio 0-4 hrs | 4-8 hrs | 8-20 hrs 20-24 hrs
1 0.65 Winter 25: 50 ws 50
2 Summer 25 50 100 50
3 0.65 Winter 25: 50 100 50
4 Summer 25 50 100 50
5 0.65 Winter 25 75 100 75
6 Summer 25 75 100 75
Closed Loop Control
Under this control strategy, a feedback information flow from the response variable is used to
decide the control variable value. Two categories of feedback control strategy are used in this
investigation: (1) Output feedback control and (2) Output integrated feedback control [Cusum
control].
Table 4: Effluent Quality under Time-Varying C ontrol Strategy
Pol | Sludge Season Aerator-in-use (hp) Effluent Quality Aerator
icy | Recirculation (kg/m*) Energy
No. | Ratio (hp-hr)
[10- 200hr]
0-4 4-8 | 8-20 | 20-24 Mean Standard
hrs hrs | hrs hrs Deviation
1 [0.65 Winter | 25 50 | 75 50 0.03083 | 0.00499 5800
2 Summer | 25 50 | 100 | 50 0.0316 0.00513 5800
3 | 0.65 Winter | 25 50 | 100 | 50 0.03045 | 0.00469 7000
4 Summer | 25 50 | 100 | 50 0.03123 | 0.00481 7000
5 | 0.65 Winter | 25 75 |100 | 75 0.03021 | 0.00476 7900
6 Summer | 25 | 75 | 100 | 75 0.03095 | 0.00489 | 7900
The strategy of basing the
sludge recirculation policy on the information on food-to-
microorganism ratio belongs to the category of state feedback control. This category of control
has not been investigated any further here for two reasons. The first reason is that this strategy
has been already studied in great detail and presented earlier (Das et al. 1995). The second
reason is that the time to measure the biomass concentration and the substrate concentration in the
aeration tank usually takes 12 hours (considering time for muffle furnace heating and cooling)
and 4 hours respectively, making regular monitoring of these variables an impractical proposition.
Output F eedback Control
Here the information on effluent quality is fed back to compare with the desired effluent quality.
The deviation from the desired value actuates the policy of aerators to be deployed. The strategy
allows an increase in aerators-in-use whenever effluent quality deteriorates and a decrease in
aerators-in-use whenever the quality improves. The increase in aerators-in-use is carried out in a
modular fashion, in steps of 25 hp each, with the aerator capacity ranging between 25-100 hp.
We assumed that the effluents are sampled every fourth hour to measure its quality and actuate
the control. The threshold values of effluent quality are fixed considering that its mean and
standard deviation values are around 0.03 kg/m’ and 0.005 kg/m’ respectively.
Table 5: Output Feedback C ontrol Strategy
Effluent Quality (kg/m”) Aerators-in-use (hp)
effluent quality < 0.025 25
0.025 < effluent quality < 0.03 50
0.03 < effluent quality < 0.035 75
effluent quality > 0.035 100
Taking cue from the results of open-loop control, the recirculation ratio was fixed at 0.65 in both
winter and summer. As before, the mean and standard deviation of effluent quality and the
aerator energy expended are used as the criteria to compare performance of control strategies.
The ‘slider-bar’ facility in the Powersim package has been used for changing the value of
aerators-in-use within the simulation run. Depending upon the effluent quality value for each
sample period of four hours, the value of aerator-in-use has been changed as per Table 5. The
mean and standard deviation of the effluent quality have been computed through an Excel file.
The corresponding aerator-energy consumption values were computed and are given along with
other results in Table 6. When compared with the best open-loop control policy (i.e., the time-
varying control), the closed-loop control policy performs better with regard to energy spent in
aerator with comparable values of mean and standard deviation of the effluent quality for the
summer season. In winter, however, the time-varying control gives better effluent quality
characteristics with less aerator energy compared to the feedback control.
Table 6: Effluent Quality under the Feedback C ontrol Strategy
SI No. Season Effluent Quality (kg/m*) Aerator-Energy
Mean Standard (hp-hr)
Deviation
1, Summer | 0.03051 0.0052 7300
2 Winter 0.03051 0.0052 6995
Output-Integrated F eedback Control (Cusum Control)
Integral control works on the basis of integrating or accumulating the deviations of the response
variable from its target value and using this information to design the control variable. In the
field of statistical control this idea has been utilized to develop the concept of Cusum
(Cumulative sum) chart. Figure 10 gives a schematic diagram of this form of control. The
difference from an output feedback control strategy is that the error signal is integrated here
before actuating the control.
Desired Aerators
Effluent
s Deviation
Quality . The Policy
7 Integration Transformation
Figure 10 Output-Integrated Feedback C ontrol
In the present investigation an attempt is made to use cusum control as a strategy for getting
acceptable effluent quality. The standard procedure for cusum control [Mitra, 1998] requires (1)
the knowledge of the target value of the effluent quality X., the standard deviation (o,) of effluent
quality, and (2) a specification of the extent of its deviation from the target. For the purpose of
this study the following values are adopted: X= 0.03 kg/ m’, 0, = 0.005 kg/ m’, and AX =
0.00125 kg/ m*.
The cusum control is normally implemented with the help of a V mask whose parameters (the
lead distance, d, and the angle of decision line ) are computed from the knowledge of the above-
mentioned values for an acceptable level of significance. For a level of significance of 0.05, the
lead distance, d, equals 23.96, and the angle of decision line, 0, is 7.12°.
The usual procedure to implement cusum control is to sample output values at regular intervals,
take their average, compute their deviations from the target value, accumulate the deviations as
time progresses, and plot them. The V-mask is then positioned on the chart such that its axis is
parallel to the horizontal axis and the midpoint of the vertical line of the V-mask coincides with
the last plotted cumulative sum of the deviations. If any past plotted value goes beyond either of
the two arms (decision lines) of the V-mask, the process is considered out of control.
The above-mentioned procedure is adopted in the present study in the following manner:
1. Effluent quality is sampled four times at an interval of 0.25 hour during the last hour of
every four-hour interval, after the steady state condition was obtained (i.e., after 100°
hour).
2. These effluent quality values were numerically averaged.
3. The model was simulated with pause at every fourth hour. The numerically averaged
effluent quality was tabulated.
4. The deviations of the average effluent quality from its target value were computed. The
accumulated deviations were also computed and plotted on a graph paper.
5. A V-mask, previously prepared on a paper with the computed values of parameters d and
0, was placed on the plot of the accumulated effluent quality as mentioned earlier.
6. Whenever any past plotted value of accumulated effluent quality went outside the two
arms of the V-mask, the process was considered out of control.
7. Following the usual practice in statistical process control, whenever a control action is
taken following an out-of-control point, the error corresponding to that point is set at zero
for accumulation at all later time points.
8. The slide-bar facility of the Powersim package was used to change the value of aerators-
in-use during the simulation of the model.
9. The effluent quality values were transferred to an Excel file for computation of its mean
and standard deviation.
The strategy for Cusum Control is shown in the Table 7.
Discussion on the results of Cusum C ontrol Chart:
The mean average effluent quality came out to be 0.0031057 kg/m’ with the standard deviation of
0.00592 kg/m*. The aerator energy consumption doing 100" to 200" hour also came out to be
8600 hp-hr. Neither quality-wise nor energy consumption-wise, this control strategy gave a
superior result compared to those for the time-varying open-loop control strategy. This inferior
result can be explained from the fact that the cusum control was applied in a sampled manner
after every four hours and during that time the under-controlled sewage passes though the reactor-
system. Reducing the sampling interval was however not a practical proposition for the reasons
cited earlier.
Table 7: Strategy for Cusum Control
Cusum Chart Results Action required
Process mean has shifted to a higher value, Increase the aerator-in-use by 1 step (25 hp).
indicating deterioration of the effluent quality
The process mean has shifted to a low value Decrease the aerator-in-use by | step (25 hp).
indicating improvement of effluent quality.
The effluent quality is within the accepted No change is needed in aerators-in-use.
range
A Comparison of the Best C ontrol Strategies
Table 8 and Table 9 give the results for the best control strategies obtained so far for winter and
summer respectively. The winter results (Table 8) indicate that as far as the aerator energy is
considered, time-varying control gives the best result. But the same thing cannot be said about
this strategy when effluent quality is considered. Strategies 1, 2 and 3 are the candidates.
Because of the extremely high aerator energy requirement, the strategy of constant control is
ignored here. A statistical test of hypothesis is done here to compare the mean effluent quality for
the time-varying control strategy with that for the output feedback control. The Z-statistic is
evaluated and compared with the Zo 9; value obtained from the standard normal table. The
hypotheses selected are: Ho: [2 = }3, Hi: 2 > 3. The test statistic values were obtained as: Zp =
2162.88 and Zo o;= 1.645. Thus the null hypothesis is rejected at a 5% level of significance and it
is inferred that as far as effluent quality is concerned, output feedback control is the best strategy.
Table 9 (for summer) indicates that output feedback control is the best if the aerator energy
requirement is considered. It is also the best if only the mean effluent quality is considered. But
when one considers the standard deviation of the effluent quality the time-varying control
strategy appears to be a candidate. The test of hypothesis is carried out to get a clear picture: Ho:
He = pb, Hi: po > ps, Zo = 123.34 and Zoos = 1.645. Thus the null hypothesis is rejected. It is
inferred that the output feedback control gives the best result considering not only the aerator
energy expended but also the effluent quality.
Table 8: Strategy-wise C omparison of the Results (Winter)
Control strategy Mean Standard Aerator Energy
Deviation in use
1. Constant Control (RR = 0.65 and 0.03009 0.00507 10000
Aerator-in-use = 100 hp)
2. Time Varying Control 0.03083 0.00499 5800
3. Output feedback control 0.03051 0.0052 6995
4. Cusum Control 0.031057 0.00592 8600
In conclusion, it can be stated that output feedback control is the most potent strategy for
activated sludge treatment systems during both winter and summer.
Table 9: Strategy-wise Comparison of the Results (Summer)
Control strategy Mean Standard Aerator Energy
Deviation in use hp-hr.
1. Constant Control 0.03074 0.00522 10000
2. Time-varying Control 0.03095 0.00489 7900
3. Output feedback control 0.03051 0.0052 7300
4. Cusum Control 0.031125 0.00573 8450
Sensitivity Analysis for the Best Strategy
Although sensitivity tests can be carried out very extensively for changes in the values of many
model variables, in what is given below only the variable sensitivity results are cited. The
variables selected are the two major variables associated with the inflows: (1) Average Inflow
Rate and (2) Inflow Substrate Concentration. Values of these variables are changed one by one
and the best policy is run. Table 10 gives the best policy results for the following:
1. The original inflow conditions (i.e., average inflow rate = 200 m/hr, and inflow
concentration = 0.15 kg/m*) are maintained.
2. Average inflow rate is reduced to 150 m’/hr, with other variables remaining at their base
values.
3. Inflow substrate concentration is increased to 0.20 kg/m’, with other variables remaining
at their base values.
Reduction of inflow rate by 25% improves the value of mean effluent quality by about 10% with
reduction in its standard deviation. Simultaneously, aerator energy expended reduces by nearly
50%. Such an improvement is expected and improves the credibility of the policy.
In the second policy sensitivity test, however, when the inflow substrate concentration is raised
by 33%, the mean effluent quality deteriorates by about 23%, with a fall in aerator energy
expense by about 12%. In spite of the use of the feedback control strategy, the effluent quality
has deteriorated due to a limitation on imposed maximum aerator capacity of 100 hp.
It only means that the initial design of limiting the aerator capacity was not correct. It is expected
that if the maximum aerator capacity value were increased, and the feedback control strategy was
designed accordingly, the results would be very acceptable. The maximum aerator capacity is
increased to 125 hp and thereafter to 150 hp. It is observed that when the model is run with the
same control strategy and with 150 hp as max aerator-in-use, the effluent quality improves (with a
reduction of its mean value to 0.030532 and a standard deviation of 0.00616) but the total aerator
energy increases to 12900 hp-hr.
It is therefore concluded that the output feedback control strategy holds good for higher influent
substrate concentration value with high.
Table 10: Sensitivity Analysis for the Best Strategy
Effluent Quality (kg/m*) Aeration Energy (hp-hr)
Mean Standard (Between 100-200 hr)
Deviation
Best Policy Average Inflow Rate
= 200 m*/hr 0.03051 0.0052 9800
Inflow Sub Case = 0.15 m*/hr
Average Inflow Rate = 150 m*/hr
Other parameters at base values 0.027164 0.0044763 5000
Inflow Substrate Concentration =
0.20 kg/m? 0.037533 0.0065711 8600
Other parameters at base values
References
Anderson, J. S., Kim. H., Me Avoy, T. J., and Hao, O. J., (2000); “Control of an Alternating
Aerobic-anoxic Activated Sludge System-I; Development of a Linearization-based Modeling
Approach” Control Engineering Practice, 28(3), pp. 271-278.
Arceivala, S. J. (1981); “Waste Water Treatment and Disposal: Engineering and Ecology in
Pollution Control”, New York: Marcel Dekker, Inc.
Bliss, P. J., and Barnas D., (1986); “Modeling Nitrification in Plant Scale Activated Sludge”,
Water Science and Technology, 18, pp. 139-148.
Clemson, B., Tang, Y., Pyne, J., and Unal, R., (1995); “Efficient Methods of Sensitivity
Analysis”, System dynamics Review, 11(1), pp. 31-49.
Das, B. K., Bandyopadhyay, M., and Mohapatra, P. K. J., (1995); “System Dynamics Modeling
of an Activated Sludge Plant”, Proceedings of 1995 International System Dynamics Conference
held at Tokyo, Japan.
Das, B. K., Bandyopadhyay, M., and Mohapatra, P. K. J., (1997); “System Dynamics Modeling
of Biological Reactors for Wastewater Treatment”, J. of Environmental Systems, 25(3), pp. 213-
240.
Das, B. K. (2003) THE PHD THESIS
Ford, D. L., and Eckenfelder, W. W., Jr., (1967); “Effects of Process Variables on Sludge Flow
Formation and Settling Characteristics”, J. of Water Pollution Control Federation, 39, pp. 1850-
1859.
Grady, C. P. L., Jr, and Lim, H. C. (1986); “Biological Waste Treatment: Theory and
Applications”, New York: Marcel Dekker, Inc.
Grady, Jr., C. P. L., and Roper, Jr., R. E., (1974); “A Model for Bio-oxidation Process, which
Incorporates the Viability Concept”, Water Research, 8, pp. 471-483.
Lawrence, A. W. and P. L. McCarty, (1970); “Unified Basis of Biological Treatment Design and
Operation”, J. of Sanitary Engineering Division, American Society of Civil Engineers, 96, pp.
757-178.
Mitra, A., (1998); “Fundamentals of Quality Control and Improvement”, Prentice-Hall
International, Inc., New Jersey, International Edition.
Monod, J. (1949); “The Growth of Bacterial Cultures”, Annual Review of Microbiology, 3, pp.
371-394,
Novak, J. T., (1974); “Temperature-Substrate Interactions in Biological Treatment” J. of Water
Pollution Control Federation, 46, pp. 1984-1994.
Picioreanu, C., Van Loosdrecht, M. C. M., and Heijnen, J. J., (1997); “Modeling the effect of
Oxygen Concentration on Nitrite Accumulation in a Bio-Film Airlift Suspension Reactor”,
Water Science and Technology, 36 (1), pp. 147-156.
Roper, Jr., R. E., and Grady, Jr., C.P.L, (1978); “A Simple Effective Technique for Controlling
Solids Retention Time in Activated Sludge Plants”, J. of Water Pollution Control Federation, 50,
pp. 702-708.
Sincero, A. P. and Sincero G. A. (1996); “Environmental Engineering - A Design Approach”,
New Jersey, USA; Prentice-Hall, Inc.
Smeers, Y., and Tyteca, D., (1984); “A Geometric Programming Model for the Optimal Design
of Wastewater Treatment Plants”, Operation Research, 32(2), pp. 314-342.
Tang, C., Brill, E.D., Jr., and Pfeffer, J. T. (1987); “Comprehensive Model of Activated Sludge
Wastewater Treatment System”; J. of Environmental Engineering Division, Proceedings of
American Society of Civil Engineers, 113(5), pp. 952-969.
Uber, J. G., Brill, E. D., Jr., and Pfeffer, J. T., (1985); “A Non-linear Programming Model of
Wastewater Treatment System: Sensitivity Analysis and a Robustness Constraint” Research
Report No. 196, Water Resources Center, University of Illinois at Urbana-Champaign.
Zhao, H., Hao, O. J., and Me Avoy, T. J., (1999); “Approaches to Modeling Nutrient Dynamics:
ASM 2, Simplified Model and Neural Nets”, Water Science and Technology, 40(2), pp. 227-234.
CC BY-NC-SA 4.0