See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260715946
Ammonia-Based Feedforward and Feedback Aeration Control in Activated
Sludge Processes
Article in Water Environment Research · March 2014
DOI: 10.2175/106143013X13596524516987 · Source: PubMed
CITATIONS
READS
57
2,748
4 authors, including:
Leiv Rieger
inCTRL Solutions Inc.
Peter Dold
119 PUBLICATIONS 3,239 CITATIONS
132 PUBLICATIONS 2,508 CITATIONS
SEE PROFILE
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
HSG Simulation View project
Mainstream Partial Nitritation Anammox View project
All content following this page was uploaded by Leiv Rieger on 08 January 2014.
The user has requested enhancement of the downloaded file.
Ammonia-Based Feedforward and Feedback
Aeration Control in Activated Sludge Processes
Leiv Rieger1*, Richard M. Jones1, Peter L. Dold1, Charles B. Bott2
ABSTRACT: Aeration control at wastewater treatment plants based
on ammonia as the controlled variable is applied for one of two reasons:
(1) to reduce aeration costs, or (2) to reduce peaks in effluent ammonia.
Aeration limitation has proven to result in significant energy savings,
may reduce external carbon addition, and can improve denitrification
and biological phosphorus (bio-P) performance. Ammonia control for
limiting aeration has been based mainly on feedback control to constrain
complete nitrification by maintaining approximately one to two
milligrams of nitrogen per liter of ammonia in the effluent. Increased
attention has been given to feedforward ammonia control, where
aeration control is based on monitoring influent ammonia load.
Typically, the intent is to anticipate the impact of sudden load changes,
and thereby reduce effluent ammonia peaks. This paper evaluates the
fundamentals of ammonia control with a primary focus on feedforward
control concepts. A case study discussion is presented that reviews
different ammonia-based control approaches. In most instances,
feedback control meets the objectives for both aeration limitation and
containment of effluent ammonia peaks. Feedforward control, applied
specifically for switching aeration on or off in swing zones, can be
beneficial when the plant encounters particularly unusual influent
disturbances. Water Environ. Res., 86, 63 (2014).
KEYWORDS: aeration control, ammonia-based control, feedforward
control, feedback control, energy savings, wastewater treatment.
doi:10.2175/106143013X13596524516987
Introduction
Ammonia-based aeration control of the activated sludge
process can lead to significant aeration energy savings, and
potential performance improvements for nitrogen and biological
phosphorus (bio-P) removal plants. Several studies report that
ammonia control leads to energy savings in the range of 15 to
25%, and to significant increases in nitrogen removal. Table 1
lists data from selected ammonia control case studies. An
extensive list is provided in a recent review of aeration control at
water resource recovery facilities (WRRFs) (Åmand et al., 2013).
Recently, ammonia feedforward control has received much
attention for optimal aeration of activated sludge systems (e.g.,
Vrečko et al., 2011; Sunner et al., 2009; Liu et al., 2005; Walz et
al., 2009; Yong et al., 2005; Shen et al., 2011). This paper aims to
clarify some of the misconceptions about this control concept.
The first step is to separate (1) the advantages and disadvantages
of ammonia control from (2) how to best implement ammonia
1
* EnviroSim Associates Ltd., McMaster Innovation Park, 175
Longwood Rd S, Suite 114A, Hamilton, Ontario, L8P 0A1, Canada;
e-mail: rieger@inCTRL.ca.
2
Hampton Roads Sanitation District, Virginia Beach, Virginia.
January 2014
control; that is, a discussion of feedforward versus feedback
control concepts.
The paper begins with an overview of the underlying reasons
for applying ammonia control including consideration of the
nitrification process and aeration requirements, and potential
causes of ammonia peaks. This is followed by a general
discussion regarding the goals of feedforward control and the
inherent dangers. A section on ammonia feedforward control
examines the advantages and disadvantages of this control
concept, and includes an overview of ammonia feedforward
controller types. Situations are identified that reveal when
feedforward control is beneficial, and when it is not beneficial.
A case study of the Hampton Roads Sanitation District’s
Nansemond WRRF is presented to illustrate the impact of
different ammonia-based control strategies, including feedback
and feedforward concepts.
Nitrification and Effluent Ammonia. Nitrification is implemented in many activated sludge plants to maintain an effluent
ammonia concentration lower than a permit level (e.g., 3 mgN/
L), and often as a precursor to denitrification where there is an
effluent total nitrogen limit. Effluent limits are typically applied
because residual ammonia may cause a critical oxygen deficit in
the receiving water, potentially resulting in harm to the
environment. Additionally, at high pH levels and high temperatures, the ammonium/ammonia equilibrium favors a greater
free ammonia concentration, which may be toxic for aquatic and
marine biota.
Conversion of ammonia to nitrate by a nitrifying biomass
requires sufficient dissolved oxygen, ammonia as substrate, a
number of nutrients, and a sufficiently long aerobic sludge
retention time (SRT) to avoid washout of nitrifiers. Nitrification
is a two-step process: (1) ammonia oxidizing bacteria (AOB)
convert ammonia to nitrite, and (2) nitrite oxidizing bacteria
(NOB) oxidize nitrite to nitrate.
In most nitrification plants, an objective is to maintain
dissolved oxygen at approximately 2 mg/L, typically resulting in
complete nitrification with low residual effluent ammonia which
is often substantially lower than the permit level. In many
instances (higher temperatures and long SRT), the effluent
ammonia may be in the range of 0.1 to 0.2 mgN/L over a one day
period. However, nitrifying plants are typically based on plug
flow designs; more often than not, the effluent shows some
breakthrough of ammonia at the high influent load time of day,
even if dissolved oxygen levels are maintained at roughly 2 mg/L.
Figure 1 shows a typical ammonia profile over 48 hours at the
aeration tank outlet and in the secondary effluent from a fully
nitrifying plant. Effluent ammonia spikes may be critical for
plants with ‘‘never-to-exceed’’ permit limits. The peak in
63
Rieger et al.
Table 1—Selected ammonia-control case studies.
Reduction
in air flow
requirements1
(%)
30
16
515
19 2
1520
20
15
20
17
1
2
Increase
in total
nitrogen
removal
(%)
Decrease in
external
carbon
(%)
50
30–50
100
36
(2 mgN/L in effluent)
50
40
40
Reference
Nielsen and Önnerth, 1995
Husmann et al., 1998
Ingildsen et al., 2002
Liu et al., 2005
Ayesa et al., 2006
Sunner et al., 2009
Walz et al., 2009
Rieger et al., 2012a
Rieger et al., 2012a
The reference for all case studies is dissolved oxygen control.
In addition, internal recycle control has been introduced.
effluent ammonia is an expected kinetic-stoichiometric consequence.
It is important to recognize that the mass of nitrifiers in the
system is set by the average ammonia load (i.e., the mass of
ammonia converted to nitrate) and the SRT. The nitrifier mass
changes slowly from day to day. Under the diurnal flow and load
patterns, the ammonia loading rate may fluctuate substantially
during the day; however, the nitrifier mass is essentially constant.
Although the mass of nitrifying organisms may remove
ammonia at the maximum rate during the high load time of
day, frequently not all of the ammonia load will be nitrified
resulting in the breakthrough of ammonia in the effluent. This is
a kinetic limitation and the situation is exacerbated if one, or
both, of the following conditions are true: (1) the plant receives
shock ammonia loads, (2) the aeration system cannot maintain
sufficient dissolved oxygen levels in all parts of the aerated
zones.
Methods for reducing the breakthrough include increasing
overall SRT, equalizing the load, utilizing sludge storage (Yuan et
al., 2000), or augmentation concepts (e.g., Salem et al., 2002;
Krhutkovaı̀ et al., 2006). Raising dissolved oxygen higher than
1.5 to 2 mg/L will likely have no impact. When dissolved oxygen
exceeds approximately 1.5 mg/L, the nitrification rate is
essentially no longer limited by dissolved oxygen (Figure 2).
Reducing the wastage rate to increase SRT will result in a slow
change in nitrifier mass over a period of days and weeks, and the
impact on nitrification will be slow. Aside from equalizing the
load, the only option for an immediate improvement in
nitrification to reduce the effluent peak is to increase the
aerated SRT (e.g., switch on aeration in swing zones at
appropriate times).
A first step in evaluating ammonia control for reducing
effluent peaks should be to determine the causes and extent of
input ammonia peaks. A detailed analysis is given in Gujer and
Erni (1978). Frequently, ammonia peaks have operational causes
(e.g., non-optimal reject water dosage or other internal recycles).
These operation-induced problems might be addressed more
easily than utilization of a complex control system.
Typically, ammonia peaks are more probable in small
catchments; however, payback of additional instrumentation
may not be justified (Devisscher et al., 2006). Large plants
typically have large catchments and more attenuated flows and
loads (Gujer and Erni, 1978). Table 2 lists several potential
causes of ammonia peaks.
Several factors impact aeration requirements other than the
nitrification oxygen demand for autotrophic organisms that
convert ammonia to nitrate. The primary factors are:
oxygen demand from heterotrophic organisms for removal
of organic matter (carbonaceous oxygen demand);
a reduction in oxygen demand as a result of organic matter
that is oxidized using nitrite or nitrate in place of oxygen as
an electron acceptor under anoxic conditions (i.e., an
‘‘oxygen credit’’);
the load of reduced inorganic matter (e.g., sulfide or ferrous
iron) which, in most cases, should be insignificant
compared with carbonaceous and nitrification oxygen
requirements;
aeration to raise low dissolved oxygen streams to the
dissolved oxygen concentration in aerated reactors;
oxygen transfer efficiency (the amount of oxygen in the
diffused air that is transferred to the liquid to satisfy the
oxygen demand); and
air flow requirements for mixing. (This should not be
ignored because it often limits the turn-down capacity for
the entire aeration system, and constrains the potential for
aeration reduction and cost savings.)
Ammonia-Based Aeration Control. Aeration control based
on ammonia measurement is essentially applied for one of two
reasons: (1) to limit aeration (aeration is limited to allow an
elevated effluent ammonia concentration close to the permit
Figure 1—Ammonia concentration at aeration tank outlet and in secondary effluent.
64
Water Environment Research, Volume 86, Number 1
Rieger et al.
Table 2—Potential causes of ammonia peaks and ways to deal
with them.
Cause of
peak
Ammonia-rich reject water or
other recycle dosage.
Septage dosage.
Storm event pushing
ammonia-rich content of
primaries (or sewer)
through plant.
Industrial discharge.
Figure 2—Monod term for dissolved oxygen (DO) dependency of
nitrification at half-saturation constant of 0.5 mg/L.
limit [i.e., to prevent complete nitrification]; the potential
benefits include energy savings, increased denitrification, and
in some cases, improved bio-P performance.), or (2) to reduce
effluent ammonia peaks (aeration is manipulated to reduce the
extent of effluent ammonia peaks).
Ammonia-based aeration control for both objectives can be
implemented using either feedback or feedforward control. This
section provides background for the two cases, and details of the
control structures are discussed later.
Case of Limiting Aeration. The first case limits aeration to
prevent complete nitrification, and maintain effluent ammonia
concentration at a setpoint of typically 1 to 2 mgN/L. (The exact
setpoint will depend on factors such as whether the effluent
permit is based on a grab sample or a flow-weighted average
value.) The intent of this control approach is to elevate ammonia
levels by a small amount while still maintaining the effluent
ammonia permit. Figure 1 shows that for much of the day, the
effluent ammonia concentration is very low (less than 0.1 mg/L),
probably much lower than the effluent permit level. In this
situation, ammonia-based aeration control can be applied to cut
back aeration over much of the day, thereby lowering dissolved
oxygen and limiting nitrification. As compared to complete
nitrification, the reduced aeration will result in energy savings.
The low dissolved oxygen should lead to increased denitrification, and in some cases, better bio-P performance.
Over-aeration at the peak load period during the day is a
potential issue with ammonia control on its own. Effluent
ammonia may increase above the setpoint over the high load
time of day (a peak in effluent ammonia occurs for the reasons
previously outlined). The response of the ammonia controller
Internal/external
cause
Potential
solutions
Internal
Equalize reject water.
Internal
External
Equalize.
Storm tanks to reduce
hydraulic overload;
bypass parts of plant.
External
Source control, equalize,
pre-treatment on site.
would be to increase aeration. However, in this situation
nitrification performance is limited by the nitrifier mass;
increasing dissolved oxygen above 1.5 mg/L will have no impact.
For this reason, ammonia controllers typically should be used in
conjunction with dissolved oxygen limits to prevent overaeration at high ammonia concentrations. The ammonia
controller can resume control authority when the peak has
passed.
One of two approaches is typically used for implementing
limited aeration ammonia-based control (Åmand et al., 2013):
(1) a cascade controller is used where the ammonia controller
determines and changes the dissolved oxygen setpoint for a
controller adjusting air flow; or (2) a control concept is used
where the ammonia controller manipulates the air flow directly.
The first approach is generally preferred because the slower
ammonia control loop manipulates the setpoint of the faster
dissolved oxygen control loop. This simplifies control system
tuning. In addition, with this approach the ammonia part can be
switched off (or even fail); the dissolved oxygen controller will
then serve as a stand-alone fall-back strategy. Because the
dissolved oxygen controller is active at all times, adequate
process control should be achieved, even if the system reverts to
a fall-back situation. Establishing a fall-back strategy with direct
ammonia control of air flow is more difficult.
In the second approach of direct ammonia control, a dissolved
oxygen sensor should still be installed to limit the air flow when
the dissolved oxygen is above a defined concentration (e.g., 2
mg/L). This will prevent over-aeration at high ammonia load.
Therefore, both control concepts require monitoring of
ammonia and dissolved oxygen to avoid a difference in sensor
costs.
Figure 3—Schematic representation of feedback controller.
January 2014
65
Rieger et al.
Figure 4—Schematic representation of feedforward controller.
When implementing direct ammonia control, special care
should be taken when tuning the control loops, because changes
in dissolved oxygen concentration occur much faster than
changes in ammonia concentration. Because ammonia is the
driving control loop and dissolved oxygen is used only as a
constraint, the control system may become unstable when the
dissolved oxygen constraint applies, unless the controllers are
carefully tuned.
An advantage of direct ammonia control is that fewer air flow
changes may be necessary because of the slower concentration
changes of ammonia as compared with dissolved oxygen. This
will be favorable for the blower and associated aeration control
equipment.
Ammonia control for limiting aeration may impact a number
of factors in plant performance:
66
For typical municipal nitrifying plants, ammonia control
reduces aeration, although should have no impact on
biochemical oxygen demand removal. The carbonaceous
oxygen requirements will be satisfied if complete (or near
complete) nitrification is achieved.
A further quoted advantage of ammonia control is that
plant performance becomes more stable. In plug flow
nitrifying plants, there is often a sudden drop in oxygen
demand at the point in the bioreactor where ammonia
removal by nitrification is complete. The position of this
point moves over the day, depending on the influent
loading rate and temperature variations, making aeration
control more difficult. Despite multiple dissolved oxygen
control loops, it is difficult to account for this step change
in oxygen demand. Depending on the location of the
dissolved oxygen probes, this may lead to over-aeration of
the downstream sections of the aerated bioreactors, even if
diffuser tapering partly accounts for the oxygen demand
profile. When using ammonia control, nitrification is never
complete, and therefore, this problem is avoided.
Increased denitrification occurs through simultaneous
nitrification and denitrification at low dissolved oxygen
concentrations (Münch et al., 1996; Jimenez et al., 2010);
conventional denitrification occurs if aeration is switched
off completely.
It has been suggested that benefits may be gained from a
denitrification shortcut where nitrite generated by AOB is
reduced directly to nitrogen gas (also called ‘‘nitrite shunt’’)
(Marcelino et al., 2011). This would lead to significant
savings in aeration energy as compared to nitrifying all the
way to nitrate, and reduced carbon needs for denitrification. However, elevated nitrite levels may lead to increased
nitrous oxide emissions (Houweling et al., 2011; Kitamura
et al., 2012). A recent study has demonstrated control of
nitrous oxide emissions in a system using the nitrite shunt
(Lemaire et al., 2011).
Sustained low dissolved oxygen levels may cause accumulation of nitrite, thereby adversely impacting effluent
disinfection, possibly reducing phosphorus uptake by
phosphate-accumulating organisms, or even triggering
sludge bulking and foaming issues. However, in contrast
to studies with constant low dissolved oxygen (e.g. Martins
et al., 2004), there is little evidence from ammonia-based
aeration control studies to suggest that these are significant
issues (Ingildsen et al., 2002; Rieger et al., 2012a).
Case of Reducing Effluent Ammonia Peaks. The second case
for ammonia-based aeration control is the perception that it
offers a means to reduce the extent of effluent ammonia peaks.
However, as previously noted, increasing the bioreactor dissolved oxygen above approximately 1.5 mg/L during a peak may
have little effect because the nitrification rate is kinetically
limited by the nitrifier mass.
Essentially, it is a misconception that the aeration rate can be
adjusted to reduce peaks in effluent ammonia. However, in
certain cases, increasing dissolved oxygen may improve nitrification performance. For example, in systems with poor diffuser
tapering, increasing aeration may raise dissolved oxygen in parts
of the bioreactor where dissolved oxygen was limiting nitrification. However, this will cause over-aeration in other parts of the
bioreactor. Nevertheless, it should be conceded that a special
case exists when ammonia-based control is applied to maintain
an elevated effluent ammonia concentration of approximately 1
to 3 mg/L. Figure 1 illustrates the effluent ammonia response
(the ammonia peaks at 2 mg/L at the high load time). If aeration
was limited to maintain ammonia at 2 mg/L between 12:00 a.m.
and 12:00 p.m. rather than nitrifying completely, the maximum
ammonia concentration at the peak would probably be in the
range of 2 to 4 mg/L, and likely closer to 2 rather than 4 mg/L.
This will depend on many factors such as the form of the
influent loading pattern and the degree of plug flow in the
aeration tanks. Additionally, when ammonia begins to increase,
the controller will raise dissolved oxygen toward the upper limit,
which will reduce the peak. For this ammonia control situation,
and any high load condition, raising the bioreactor dissolved
oxygen prior to the peak load and decreasing the ammonia
concentration will provide a degree of buffer before the peak
occurs. However, the extent of the buffering action depends on
many factors and can only be evaluated through dynamic
simulation.
Essentially, the only strategy for significantly reducing effluent
ammonia peaks without adding additional reactor volume (for
sludge storage strategies see Yuan et al., 1998, 2000; for
Water Environment Research, Volume 86, Number 1
Rieger et al.
Figure 5—Planning, implementation, and operation phases of advanced aeration control systems.
augmentation strategies see Salem et al., 2002 and Krhutkovaı̀ et
al., 2006) is to increase the aerated sludge mass in systems that
include swing zones. An ammonia controller can be used to
switch aeration on or off in the swing zone. This approach will
have an immediate impact on nitrification performance by
increasing the aerated fraction of the nitrifier mass. (A special
case applies to oxidation ditches, where aerobic and anoxic
zones are not clearly defined, and changes in aeration intensity
will have an impact on the size of aerated zones.)
As noted, ammonia-based aeration control for either of the
two objectives (limiting aeration or reducing ammonia peaks)
can be implemented using either feedback or feedforward
control, or a combination of both. In each case the goal is the
same, which is to maintain a desired (setpoint/reference)
ammonia concentration. The following section presents the
principles of feedback and feedforward control, followed by a
critical discussion about the application of ammonia-based
feedforward concepts for aeration control.
Feedback versus Feedforward Control. Feedback control
involves a measured process variable (controlled variable) that is
input to the controller (e.g., bioreactor dissolved oxygen
concentration). Control action is based on the difference (error)
between the measured value and the desired value (setpoint/
reference). The objective is to reduce the error by the control
action (e.g., by adjusting the air flow). Figure 3 shows a
schematic of a feedback control loop.
January 2014
A feedforward controller measures a process disturbance (e.g.,
influent ammonia load) and uses a model to predict the behavior
of the controlled system. The predictions are then used to
calculate the control action to be taken (e.g., changing air flow or
a dissolved oxygen setpoint). Again, the objective is to maintain
the controlled variable at the setpoint/reference value; however,
the controlled variable is not measured. Figure 4 shows a
schematic representation of a feedforward controller.
Control Structure Selection. Typically, a feedback control
structure is preferable over a feedforward structure, because
feedback control is based on the measured output of the process
and does not require a model of the controlled system. A
feedforward controller requires a model of the controlled system
(Figure 4); therefore, it requires knowing the impact of potential
disturbances and many other factors (e.g., ongoing bio-chemical
and physico-chemical processes, physical dimensions, operational settings and hydraulics, and impacts of disturbances such
as influent load and temperature).
A possible criticism of feedback control is that an error must
exist before a control action can be taken. In systems with short
retention times, it may be too late to accommodate sudden
disturbances because of long sensor or actuator response times,
or slow reactions to control actions.
A common argument for feedforward control is that potential
disturbances are addressed before they actually disturb the plant.
For example, if a peak is monitored upstream of the process unit,
67
Rieger et al.
Figure 6—Model configuration of the Nansemond water resource recovery facility.
adequate actions are potentially triggered to prepare the plant
for the increased load (or flow). The intent of the feedforward
controller is to use a time advantage to react more quickly to an
influent disturbance which may otherwise exceed the capacity of
the plant, resulting in an exceedance of effluent ammonia. The
success of a feedforward controller is therefore measured in
terms of its ability to accommodate disturbances and reduce
impacts on effluent quality.
In wastewater treatment, feedforward concepts are successfully applied in cases where a fast reaction to disturbances is
required, and where the impact of the control action is simple to
predict. Examples are flow control, dosage of precipitants for
chemical phosphorus removal, or addition of external carbon
based on an upstream signal. In such cases, using feedback
concepts may lead to violation of effluent permits or oscillating
controlled variables, because the measured signal is too late to
compensate for the disturbances.
Inherent Dangers of Feedforward Control. Feedforward control
requires a potentially simple model to predict response in the
downstream controlled system. A perfect feedforward controller
requires a perfect model, because there is no information
available to the controller regarding how the system responds to
controller actions and disturbances, such as changing loadings,
temperature, and SRT. Therefore, most feedforward controllers
use a downstream signal to either adjust the model or trim the
control action by feedback control, thereby accounting for
deficiencies in the model. (Limits are also typically imposed on
the control action; for example, a maximum dissolved oxygen
concentration to avoid the over-aeration issue.)
When applying feedforward control, special attention should
be given to the inherent risk of prediction model failure and the
potential violation of effluent limits.
Disadvantages of Feedforward Controllers. A very simple model
needs less input data but will have limited capability to predict
how the process reacts to disturbances and controller actions.
Although a complex model may be more accurate in its
predictions, it will require several sensors as inputs. Therefore,
as compared to feedback control, in most cases feedforward
control concepts (1) need more sensors, (2) are more complex, (3)
need increased maintenance/controller tuning, and (4) place an
additional burden on operators.
Selection Criteria. The preferred control structure depends on
many factors including the influent load and temperature
variations, the specific plant configuration, the response time
of sensors, actuators and equipment such as blowers, and the
68
effluent permit (limits and averaging time). When testing control
strategies, one criterion is the ‘‘control authority’’ (Olsson and
Newell, 1999), which is the ability of the controllers to maintain
a given setpoint or track a trajectory of setpoints over a range of
conditions (both increasing and decreasing disturbances).
Regarding feedback versus feedforward control, the issue is
whether the feedback controller can react within a reasonable
time to maintain the controlled variable within an acceptable
range around the setpoint. If the downstream signal is too late,
and an earlier signal would be advantageous, the measurement
point can be moved upstream (e.g., from effluent to last
bioreactor). Feedforward control should only be used if: (1)
the process dynamics are slow relative to the frequency and
amplitude of the disturbances, (2) feedforward control actions
have sufficient control authority and are able to reduce the
impact of a disturbance, (3) the controlled system with potential
disturbances can be described in a model with sufficient
accuracy, and (4) the disturbances can be measured or reliably
estimated on-line.
Ammonia-Based Feedforward Control. This section focuses
on a feedforward control structure for the two cases of aeration
control (limiting aeration and reducing effluent ammonia
peaks). Advantages and possible pitfalls of feedforward ammonia-based aeration control are discussed.
Ammonia Feedforward Control for Limiting Aeration. The
intent of a feedforward controller is to use a time advantage to
react rapidly to an influent peak, which otherwise may exceed
the capacity of the plant. If the control goal is to limit aeration,
and thereby save energy and improve denitrification, feedforward concepts may not be the best choice for the following
reasons:
Effluent ammonia concentration typically changes slowly in
activated sludge plants because the influent load disturbances are usually slow and disturbances are attenuated
through the long retention time process. In most instances,
ammonia feedback control is sufficiently fast and provides
acceptable control authority to maintain an ammonia
setpoint within reasonable bounds. Introducing feedforward control at these plants would not provide any
additional benefits.
Feedforward control should include safety factors against
model prediction inaccuracies. Therefore, the potential cost
savings and other benefits from ammonia control may not
be fully realized.
Water Environment Research, Volume 86, Number 1
Rieger et al.
Table 3—Selected control strategies tested in Rieger et al. (2012b).
Control strategy
Strategy 1 (base case): dissolved oxygen control.
Strategy 2b: NH4 feedback (FB).
Strategy 4: NH4 feedforward (FF) aeration intensity plus FB.
Strategy 5 (new scenario): NH4 FF aerated volume plus FB.
1
Description
Existing strategy: dissolved oxygen control to fixed setpoints in three aeration
zones.
Ammonia feedback control (PID1) changing the dissolved oxygen setpoints
between 02 mg/L.
Feedforward/feedback ammonia aeration intensity control changing dissolved
oxygen setpoints between 02.5 mg/L. Selection of active control loop based
on loop requesting higher dissolved oxygen.
Feedforward ammonia control switching swing zone aeration on/off (aerated
volume control); feedback ammonia control changing dissolved oxygen
setpoint between 02 mg/L.
PID controller ¼ proportional-integral-derivative controller.
For most plants, the lack of control authority to maintain
an ammonia setpoint as a result of ammonia load variations
is not caused by a delayed signal in a feedback control
system; rather, it is because of the limited mass of nitrifiers
in aerobic bioreactors. An early increase of aeration
intensity can provide a certain buffer capacity although is
not able to provide more nitrifiers.
Limiting aeration (reducing the aeration intensity) has a
one-sided control authority. It is effective for limiting the
nitrification rate by reducing dissolved oxygen, however, it
is not able to increase nitrification capacity. An early
control action can be used to create a buffer for incoming
ammonia load peaks, although the feedforward model must
be accurate in terms of predicted nitrification capacity and
in determining the point when increased aeration intensity
is required. Inaccurate models may result in an unnecessary
increase in energy consumption or reduced plant performance. Methods of addressing ammonia peaks are further
discussed in the following section.
Ammonia Feedforward Control for Reducing Ammonia
Effluent Peaks. Typically, feedforward concepts are applied when
effluent peaks must be reduced; in this case, effluent ammonia
peaks. A first step in evaluating ammonia control strategies is to
analyze whether feedforward control is necessary, or whether a
simple feedback controller can maintain the ammonia concentration within an acceptable error range. This section considers
different methods for implementing ammonia feedforward
control, and the resulting control authority of the proposed
control strategies.
Essentially, there are two ways to use an ammonia measurement for feedforward control in an effort to maintain an effluent
ammonia setpoint through aeration: (1) by changing the aeration
intensity (e.g., varying the dissolved oxygen setpoint or direct
ammonia control), or (2) by adding more aerated volume (e.g.,
switching on aeration in a swing zone). The first option is called
‘‘aeration intensity control’’ and the second option is called
‘‘aerated volume control’’.
‘‘Aeration intensity control’’ (first option) has both downsides
and potential benefits. A major downside of ammonia feedforward aeration intensity control is often its limited control
authority. This is different from controlling, for example,
denitrification by adding external carbon where the denitrification is essentially directly linked to carbon addition. With
ammonia aeration intensity control, increasing aeration to
January 2014
account for increased loading may help, however the nitrification capacity may be limited by the concentration of nitrifiers.
The potential benefit of an ammonia feedforward aeration
intensity controller is mainly the ability to increase aeration
sooner than a feedback controller, and therefore anticipate the
impact of a sudden load disturbance. This is achieved in one of
two ways: (1) by lowering the ammonia concentration in the
reactor to a minimum to provide dilution for the arriving
ammonia peak, or (2) by raising the air flow in anticipation of
increased oxygen demand.
In situations where the ammonia load is low, an ammonia
dissolved oxygen controller has a direct impact on the
nitrification rate (high control authority). However, the savings
potential when the ammonia load is low can likely be addressed
by using a feedback ammonia controller. This minimizes the risk
of violating effluent ammonia permits as a result of insufficient
aeration, because the target variable is measured (feedback) and
not predicted (feedforward).
‘‘Aerated volume control’’ (second option) is a more
advantageous control strategy. An ammonia volume controller
provides a higher control authority for situations in which the
incoming ammonia load is greater than the nitrification capacity
of the typically aerated reactors. By switching on aeration in
swing zones, this controller can increase the mass of active
nitrifiers. Therefore, this controller is better suited for accommodating peak loads. An additional ammonia feedback dissolved
oxygen controller will address low loading situations. This
control strategy is more beneficial because the two loading
situations are managed by two independent control loops with
high control authority for each specific situation.
Special Cases for Ammonia Feedforward Control. Feedforward
ammonia control can be useful in locations (e.g., Germany)
where peak effluent limits apply; that is, where plants are
permitted based on grab samples with a never-to-exceed limit.
Another example would be where extreme influent ammonia
peaks occur. However, before setting up a controller, a first step
would be to identify the cause of the peak, and attempt to reduce
the peak through operational changes, equalization, or pretreatment at industrial sites.
To summarize the discussion on feedforward control, special
attention should be given to the control authority of a
feedforward controller (see Nansemond WRRF case study
below). Conventional biological nurient removal plants often
have long retention times and therefore slow response times for
ammonia. Feedforward control may have benefits for systems
69
Rieger et al.
Figure 8—Comparison of ammonia concentrations in last
aerated reactor for Strategies 2b (NH4 feedback control) and 4
(NH4 feedforward plus feedback control) at 12 8C.
Figure 7—Comparison of dissolved oxygen (DO) levels for
Strategies 2b (NH4 feedback control) and 4 (NH4 feedforward
plus feedback control) at 12, 20, and 30 8C.
with a short hydraulic retention time (e.g., membrane bioreactors). Plants with ammonia controlled swing zones have a higher
control authority to address ammonia peaks. However, care
should be taken when tuning the different control loops to avoid
the creation of sudden jumps or drops in the ammonia
concentration by the feedforward volume ammonia controller
at the sensor location of the feedback control loop. These jumps
or drops could lead to a delay in the detection of sudden peaks
by the feedback controller, resulting in even higher ammonia
effluent peaks (Brischke et al., 2010).
Designing Ammonia-Based Aeration Control. Designing
ammonia-based aeration control strategies is a complex task and
requires involvement of various experts from different disciplines. Figure 5 shows a schematic view of the various phases for
planning, implementing, and operating advanced aeration
control systems. The design requires knowledge of detailed
information about the treatment plant under study. A significant
effort should be made to fully understand the plant, and all
required details and constraints should be taken into account.
Although a control strategy can be designed, implemented,
and tuned without a model, the use of a dynamic simulator in
the design process is strongly recommended. A dynamic plant
model can combine the different fields of expertise (e.g., process,
instrumentation, control, and operation), and allows off-line
testing of various control strategies without the need for a timeconsuming pilot, or full-scale testing. When evaluating different
ammonia control strategies, the model can provide valuable
results to assess the needs for feedforward control, or determine
whether feedback control would be sufficient.
70
It is essential to have a quantifiable understanding of the
desired objectives for implementing advanced control. It is often
helpful to begin the time-consuming process of instrument
testing early in the process because the operators need time to
develop confidence in the sensors; some sensors may not be the
ideal choice for the location or the planned control strategy. In
the design phase, regular interaction between control system
designers, process engineers, plant personnel, and others
involved is important for a tailored control system. A dynamic
model can provide a common language in this process.
Aspects which should not be underestimated are operator
training, and developing an incentive system for plant personnel
(Rieger and Olsson, 2012).
Case Study—Nansemond WRRF. The Nansemond WRRF
has undergone a major upgrade of its treatment capacity,
modifying it from a three-stage VIP process (Virginia Initiative
Process) to a five-stage Bardenpho process configuration to fulfill
more rigorous effluent requirements. The plant is a 115 000 m3/d
(114 ML/d) facility currently operating at approximately 60 000
m3/d (61 ML/d) (basis of modeling) with a relatively large
geographic collection system and an industrial contribution of
approximately 20 percent. An extensive simulation project has
been performed to identify optimal aeration control strategies
that reduce aeration energy, and optimize the dosage of external
carbon (Rieger et al., 2012b). The Nansemond plant utilizes
anaerobic digestion, centrate equalization, and centrate treatment
by the Ostara struvite recovery process. The basis for plant
modeling was anaerobic digestion and full centrate equalization.
Figure 6 presents the plant model configuration that was used in
the simulation project. One objective was to analyze the need for,
and the potential benefits of ammonia feedforward control.
Table 3 lists some of the control strategies tested. Each
strategy was evaluated for temperatures of 12, 20, and 30 8C, and
involved simulating plant performance subject to typical influent
flow and load variations over one week. Additional information
on the control scenarios can be found in Rieger et al. (2012b).
Control Authority of Aeration Intensity Controller. The purpose
of a feedforward control strategy with the potential to reduce
effluent ammonia peaks is to increase aeration intensity by
increasing dissolved oxygen setpoints when the incoming
ammonia load is higher than a predicted nitrification capacity.
The method used by the feedforward model to estimate the
nitrification capacity is described in Rieger et al. (2012a). As
shown in Figure 7, ammonia feedforward aeration intensity
control (Strategy 4) was only active at very low temperatures
(12 8C), although it was inactive at temperature scenarios of
Water Environment Research, Volume 86, Number 1
Rieger et al.
Figure 9—Artificial total Kjeldahl nitrogen (TKN) influent peak on
top of normal influent pattern.
20 and 30 8C. When active, the ammonia feedforward controller
had a very limited impact on effluent concentrations under
normal load conditions with a peak reduction of less than 0.1
mgN/L (Figure 8).
Because only minimal data on diurnal ammonia influent
variations and potential peak loads were available, an artificial
ammonia peak as a ‘‘worst case’’ scenario was created to analyze
the feedforward controller performance (Figure 9). Figure 10
shows the results comparing the effluent concentrations for
Strategies 2b (NH4 feedback) and 4 (NH4 feedforward plus
feedback changing dissolved oxygen setpoint). The feedforward
controller only achieved limited control authority over ammonia
peak loadings; the difference between the resulting ammonia
effluent concentrations was less than 0.5 mg/L.
Control Authority of Aerated Volume Controller. A second
feedforward control strategy with improved control authority to
reduce effluent ammonia peaks is to switch on aeration in a
swing zone when the incoming ammonia load is higher than a
predicted nitrification capacity (i.e., aerated volume control).
Details of the feedforward model are described in Rieger et al.
(2012a). Switching on aeration in swing zones increases the
aerated SRT, and consequently increases the mass of active
nitrifiers in the system. Figure 11 shows the dissolved oxygen
concentration at the end of the aeration tank for Strategies 2b
and 4. The dissolved oxygen concentration in Swing zone
AAA_C is shown for Strategy 5 (NH4 feedforward volume
control plus NH4 feedback dissolved oxygen control). It is
evident that aeration is switched on in the swing zones when the
predicted nitrification capacity is below the measured incoming
ammonia load. Figure 12 compares the ammonia concentrations
Figure 10—Comparison of ammonia and dissolved oxygen (DO)
concentrations at end of aerated zones (Aer4-7_fg) when subject
to (artificial) ammonia peak for Strategies 2b (NH4 feedback
control) and 4 (NH4 feedforward plus feedback control) at 12 8C.
January 2014
Figure 11—Comparison of different ammonia-based control
strategies at 12 8C (FB ¼ feedback; FF ¼ feedforward; DO ¼
dissolved oxygen).
in the last zone of the aeration tanks; Figure 13 shows the same
for the effluent. At this point, the feedforward controller has
sufficient control authority to maintain the ammonia concentration below 2 mg N/L in the last aerated tank, including
situations of high loading. The effluent ammonia concentrations
for Strategy 5 are constant at approximately 0.4 mg N/L.
Strategy 4 has minimal impact on effluent ammonia concentrations as compared to Strategy 2b, which has only a feedback
controller.
Figure 14 shows a comparison of energy consumption and
Figure 15 shows the carbon dosage for the four control
strategies. The difference in aeration costs between the two
feedforward strategies is marginal. However, Strategy 5 shows a
methanol demand that is approximately 510 L/d greater than for
Strategy 4. This is because of the reduced pre-anoxic zone, and
consequently, lesser use of influent carbon. Assuming a cost of
$5.70 per liter, this results in an annual difference of $75 000.
Additional information on influent loads and plant operation is
provided in Rieger et al. (2012b).
Conclusions
Aeration control based on ammonia measurement is applied
for one of two reasons:
(1) Aeration control may be used to limit aeration to reduce
operating costs, and potentially improve performance. The
approach is used to partially limit nitrification while
maintaining a target effluent ammonia concentration below
the permit value. The potential benefits include energy
savings, increased denitrification, reduced external carbon
dosage, and improved bio-P performance. Manipulating
Figure 12—Ammonia concentrations at end of aerated zones
(Aer4-7_fg) for Strategies 2b, 4, and 5 at 12 8C (FB ¼ feedback; FF
¼ feedforward; DO ¼ dissolved oxygen).
71
Rieger et al.
Figure 13—Ammonia in effluent for Strategies 2b, 4, and 5 at 12
8C (FB ¼ feedback; FF ¼ feedforward; DO ¼ dissolved oxygen).
Figure 15—Methanol dosage for Strategies 1, 2b, 4, and 5 at
12 8C.
aeration intensity has a high control authority to limit
nitrification by reducing dissolved oxygen levels, although
does not increase nitrification capacity when dissolved
oxygen exceeds approximately 1.5 mg/L.
(2) Aeration control may be used to manipulate aeration to
reduce effluent ammonia peaks. Two approaches were
evaluated in this study: (i) controlling aeration intensity,
and (ii) adjusting aerated volume fraction. The mass of
nitrifiers depends on the average ammonia load removed
and the SRT, and changes slowly from day to day. At peak
load times, the nitrification capacity of the system may be
exceeded, resulting in an effluent ammonia breakthrough.
This is a kinetic constraint and cannot be addressed by
increased aeration intensity. The only option for an
immediate improvement in the reduction of ammonia
effluent peaks by nitrification is to increase the active
nitrifier mass. This can be achieved by switching on aeration
in swing zones, or dosing stored nitrifiers (Yuan et al., 2000).
However, the latter would require additional reactor
volume.
using feedback control, suggesting that feedback control is the
more suitable approach. Typically, the ammonia concentration
in the effluent varies slowly; feedforward control rarely provides
additional advantages.
A current perception is that feedforward control is beneficial
for reducing effluent ammonia peaks. This is based on the belief
that feedforward control will provide an early warning of an
influent load increase, and aeration intensity can then be
increased to avoid an effluent peak in ammonia. However, as
previously discussed, the peaks generally result from a kinetic
constraint as a result of the limited nitrifier mass, and increasing
dissolved oxygen in aerated zones likely will not help. Essentially,
the only means for rapidly responding to influent loading peaks
and reducing ammonia breakthrough is to increase the mass of
active nitrifiers (e.g., by switching on aeration in swing zones or
dosing stored nitrifying sludge). Switching aeration on and off in
swing zones can possibly be implemented using the simpler
feedback control approach, and should not be discounted.
However, this may be the one situation in which feedforward
control offers benefits.
When feedforward control can be justified (e.g., when regular
peaks occur which cannot be equalized or in the case of neverto-exceed effluent limits), the selected control strategy should be
tested to assess the ability to reduce effluent ammonia peaks.
Careful analysis should be applied to determine whether
feedforward control offers any benefit over standard feedback
control. The case study of Nansemond WRRF showed that
feedforward control was only active at very low temperatures. In
addition, the feedforward aeration intensity control provided
minimal control authority; therefore, the additional costs would
significantly reduce potential savings as compared to feedback
ammonia control.
Whenever process control strategies are to be implemented,
dynamic simulation is an excellent tool for testing, and allows a
tailored design for a specific plant.
Submitted for publication December 4, 2012; accepted for
publication April 15, 2013.
Ammonia-based aeration control for both objectives can be
implemented using either feedback or feedforward control. In
general, if the same control objective can be achieved by either
method, feedback control is preferred because: (1) fewer sensors
are required, (2) development of a process model is unnecessary,
and (3) feedback control is more robust. An ammonia
feedforward control strategy requires at least two ammonia
probes (one upstream and one downstream), one dissolved
oxygen probe, and a flow meter. To increase accuracy of the
model, additional sensors for temperature and mixed liquor
suspended solids should be integrated. This results in a
significant investment and increased operation and maintenance
costs as compared to feedback control.
The paper demonstrates that the objective of adjusting
aeration intensity to limit nitrification can usually be achieved
References
Figure 14—Energy costs for Strategies 1, 2b, 4, and 5 at 12 8C.
72
Åmand, L.; Olsson, G.; Carlsson, B. (2013) Aeration Control—A Review.
Water Sci. Technol., 67, 2374–2398.
Ayesa, E.; De la Sota, A.; Grau, P.; Sagarna, J. M.; Salterain, A.; Suescun, J.
(2006) Supervisory Control Strategies for the New WWTP of
Galindo-Bilbao: The Long Run from the Conceptual Design to the
Full-Scale Experimental Validation. Water Sci. Technol., 53 (45),
193–201.
Brischke, K.; Olds, K.; Adams, D.; Hardison, J.; Rieger, L. (2010) Using
Whey as a Supplemental Carbon Source Under Real Time Control
Water Environment Research, Volume 86, Number 1
Rieger et al.
Conditions—Or a Story of Turds and Whey. Proceedings of the 83rd
Annual Water Environment Federation Technical Exposition and
Conference [CD-ROM]; New Orleans, Louisiana, Oct 26; Water
Environment Federation: Alexandria, Virginia.
Devisscher, M.; Ciacci, G.; Fé, L.; Benedetti, L.; Bixio, D.; Thoeye, C.; De
Gueldre, G.; Marsili-Libelli, S.; Vanrolleghem, P. A. (2006)
Estimating Costs and Benefits of Advanced Control for Wastewater
Treatment Plants—The MAgIC Methodology. Water Sci. Technol.,
53 (45), 215–223.
Gujer, W.; Erni, P. (1978) The Effect of Diurnal Ammonium Load
Variation on the Performance of Nitrifying Activated Sludge
Processes. Prog. Water Technol., 10 (56), 391–407.
Houweling, D.; Wunderlin, P.; Dold, P.; Bye, C.; Joss, A.; Siegrist, H.
(2011) N2O Emissions: Modeling the Effect of Process Configuration and Diurnal Loading Patterns. Water Environ. Res., 83, 2131–
2139.
Husmann, M.; Orth, H.; Schlegel, S.; Teichgräber, B. (1998) Application
of Process Control for Improved Nitrogen Removal. Water Sci.
Technol., 38 (3), 263–269.
Ingildsen, P.; Jeppsson, U.; Olsson, G. (2002) Dissolved Oxygen
Controller Based on On-Line Measurements of Ammonium
Combining Feed-Forward and Feedback. Water Sci. Technol., 45
(45), 453–460.
Jimenez, J.; Dursun, D.; Dold, P.; Bratby, J.; Keller, J.; Parker, D. (2010)
Simultaneous Nitrification-Denitrification to Meet Low Effluent
Nitrogen Limits: Modeling, Performance and Reliability. Proceedings of the 83rd Annual Water Environment Federation Technical
Exposition and Conference [CD-ROM]; New Orleans, Louisiana,
Oct 26; Water Environment Federation: Alexandria, Virginia.
Kitamura, K.; Nemoto, T.; Matsushita, S.; Uchida, M.; Yoda, Y.; Kassai, K.;
Hamamoto, A.; Satoh, K. (2012) Discussion about Cause of Nitrous
Oxide (N2O) Emission in Wastewater Treatment Plant, Based on
Long-Term Continuous Measurement. Proceedings of the 85th
Annual Water Environment Federation Technical Exposition and
Conference [CD-ROM]; New Orleans, Louisiana, Sept 29Oct 3;
Water Environment Federation: Alexandria, Virginia.
Krhutkovaı̀, O.; Novaı̀k, L.; Pachmanovaı̀, L.; Benaı̀kovaı̀, A.; Wanner, J.;
Kos, M. (2006) In Situ Bioaugmentation of Nitrification in the
Regeneration Zone: Practical Application and Experiences at FullScale Plants. Water Sci. Technol., 53 (12), 39–46.
Lemaire, R.; Chauzy, J.; Veuillet, F.; DiMassimo, R.; Sorensen, K.; Deleris,
S. (2011) Advanced Control System to Reduce N2O Emission and
Improve Performance of an SBR Treating N-Rich Effluent Via
Nitrite Pathway. Proceedings of the 84th Annual Water Environment
Federation Technical Exposition and Conference [CD-ROM]; Los
Angeles, California, Oct 1519; Water Environment Federation:
Alexandria, Virginia.
Liu, W.; Lee, G. J. F.; Schloth, P. E.; Serra, M. E. (2005). Side by Side
Comparison Demonstrated a 36% Increase of Nitrogen Removal
and 19% Reduction of Aeration Requirements Using a Feed
Forward Online Optimization System. Proceedings of the 78th
Annual Water Environment Federation Technical Exposition and
Conference [CD-ROM]; Washington, D.C., Oct 29Nov 2; Water
Environment Federation: Alexandria, Virginia.
January 2014
View publication stats
Marcelino, M.; Wallaert, D.; Guisasola, A.; Baeza, J. A. (2011) A TwoSludge System for Simultaneous Biological C, N and P Removal Via
the Nitrite Pathway. Water Sci. Technol., 64, 1142–1147.
Martins, A. M. P.; Pagilla, K.; Heijnen, J. J.; van Loosdrecht, M. C. M.
(2004) Filamentous Bulking Sludge—A Critical Review. Water Res.,
38, 793–817.
Münch, E. V.; Lant, P.; Keller, J. (1996) Simultaneous Nitrification and
Denitrification in Bench-Scale Sequencing Batch Reactors. Water
Res., 30 (2), 277–284.
Nielsen, M. K.; Önnerth, T. B. (1995) Improvement of a Recirculating
Plant by Introducing STAR Control. Water Sci. Technol., 31 (2),
171–180.
Olsson, G.; Newell, B. (1999) Wastewater Treatment Systems: Modelling,
Diagnosis and Control. IWA Publishing: London.
Rieger, L.; Olsson, G. (2012) Why Many Control Systems Fail. Water
Environ. Technol., 24 (6), 42–45.
Rieger, L.; Takács, I.; Siegrist, H. (2012a) Improving Nutrient Removal
While Reducing Energy Use at Three Swiss WWTPs Using
Advanced Control. Water Environ. Res., 84 (2), 171–189.
Rieger, L.; Bott, C. B.; Balzer, W. J.; Jones, R. M. (2012b) Model-Based
Aeration System Design—Case Study Nansemond WWTP. Proceedings of the 85th Annual Water Environment Federation
Technical Exposition and Conference [CD-ROM]; New Orleans,
Louisiana, Sept 29Oct 3; Water Environment Federation: Alexandria, Virginia.
Salem, S.; Berends, D.; Heijnen, J. J.; van Loosdrecht, M. C. M. (2002)
Model-Based Evaluation of a New Upgrading Concept for NRemoval. Water Sci. Technol., 45 (6), 169–176.
Shen, T.; Shi, H.; Shi, H.; Jing, H.; Xiong, H. (2011) Feedforward Control
for Nitrogen Removal in a Pilot-Scale Anaerobic-Anoxic-Oxic Plant
for Municipal Wastewater Treatment. Front. Environ. Sci. Engin.
China, 5, 130–139.
Sunner, N.; Haeck, M.; Thornton, A. (2009) Evaluation of a Full-Scale
Activated Sludge Real Time Control System. Proceedings of the 82nd
Annual Water Environment Federation Technical Exposition and
Conference [CD-ROM]; Orlando, Florida, Oct 1721; Water
Environment Federation: Alexandria, Virginia.
Vrečko, D.; Hvala, N.; Stražar, M. (2011) The Application of Model
Predictive Control of Ammonia Nitrogen in an Activated Sludge
Process. Water Sci. Technol., 64, 1115–1121.
Walz, T.; Coughenour, J. R.; Williams, K.; Jacobs, J.; Shone, L.; Stahl, T.;
Kestel, S.; Palmer, D. (2009) Energy Savings at Phoenix 23rd Avenue
Wastewater Treatment Plant Using Feed-Forward Process Control.
Proceedings of the 82nd Annual Water Environment Federation
Technical Exposition and Conference [CD-ROM]; Orlando, Florida,
Oct 1721; Water Environment Federation: Alexandria, Virginia.
Yong, M.; Yongzhen, P.; Shuying, W. (2005) Feedforward-Feedback
Control of Dissolved Oxygen Concentration in a Predenitrification
System. Bioprocess Biosyst. Eng., 27, 223–228.
Yuan, Z.; Bogaert, H.; Vansteenkiste, G.; Verstraete, W. (1998) Sludge
Storage for Countering Ammonia Shock Loads and Toxicity
Incidents. Water Sci. Technol., 37 (12), 173–180.
Yuan, Z.; Bogaert, H.; Leten, J.; Verstraete, W. (2000) Reducing the Size of
a Nitrogen Removal Activated Sludge Plant by Shortening the
Retention Time of Inert Solids Via Sludge Storage. Water Res., 34
(2), 539–549.
73