Ica Report Haimi Et Al PDF

Teknillinen korkeakoulu Vesihuoltotekniikan julkaisu
Helsinki University of Technology Water and Wastewater Engineering

Espoo 2009
ADVANCED OPERATION AND CONTROL METHODS OF MUNICIPAL

WASTEWATER TREATMENT PROCESSES IN FINLAND
Henri Haimi, Michela Mulas, Kristian Sahlstedt and Riku Vahala

Teknillinen korkeakoulu Vesihuoltotekniikan julkaisu
Helsinki University of Technology Water and Wastewater Engineering
Espoo 2009
ADVANCED OPERATION AND CONTROL METHODS OF MUNICIPAL

WASTEWATER TREATMENT PROCESSES IN FINLAND
Henri Haimi, Michela Mulas, Kristian Sahlstedt and Riku Vahala
Teknillinen korkeakoulu
Vesihuoltotekniikka
Helsinki University of Technology

Water and Wastewater Engineering
Helsinki University of Technology
Water and Wastewater Engineering
P.O. Box 5200
FI-02015 TKK
Finland
Tel. +358 9 4511
Fax. +358 9 451 3856
E-mail: contact.vesihuolto@tkk.fi
ISBN 978-951-22-9975-1
Contents
INTRODUCTION ............................................................................................................................................. 4
1.1 MOTIVATIONS ............................................................................................................................................ 4
1.2 OBJECTIVES ............................................................................................................................................... 5
1.3 REPORT OVERVIEW ..................................................................................................................................... 6
PART I............................................................................................................................................................... 7
CHAPTER 2. MODELLING AND SIMULATION.......................................................................................... 9
2.1 ACTIVATED SLUDGE PROCESS MODELS ........................................................................................................ 9
Bioreactor models....................................................................................................................................... 9
Secondary settler models........................................................................................................................... 10
ASP Simplified models .............................................................................................................................. 11
2.2 IWA/COST BENCHMARK SIMULATION MODELS ....................................................................................... 14
2.3 PLANT-WIDE MODELS ............................................................................................................................... 16
2.4 PROCESS SIMULATION ENVIRONMENT ....................................................................................................... 17
CHAPTER 3. PROCESS CONTROL............................................................................................................. 20
3.1 FEEDBACK CONTROL ALGORITHMS ............................................................................................................ 20
On-off control........................................................................................................................................... 20
PID algorithms......................................................................................................................................... 21
3.2 FEEDFORWARD CONTROL ......................................................................................................................... 22
3.3 ADVANCED CONTROL ALGORITHMS .......................................................................................................... 23
Cascade Control....................................................................................................................................... 23
Model predictive control ........................................................................................................................... 24
Fuzzy logic control ................................................................................................................................... 26
Artificial neural networks.......................................................................................................................... 28
3.4 CONTROL APPLICATIONS........................................................................................................................... 29
BSM1 applications.................................................................................................................................... 29
BSM2 applications.................................................................................................................................... 33
Real plant control examples ...................................................................................................................... 34
3.4 PLANT-WIDE CONTROL ............................................................................................................................. 36
CHAPTER 4. SENSORS AND DATA ANALYSIS ........................................................................................ 39
4.1 SENSORS AND ANALYSERS ....................................................................................................................... 39
On-line sensors......................................................................................................................................... 39
Soft-sensors .............................................................................................................................................. 39
Predicting the wastewater characteristics ................................................................................................. 41
4.2 DATA ANALYSIS AND FAULT DETECTION ................................................................................................... 42
Data screening ......................................................................................................................................... 42
Data reconciliation................................................................................................................................... 42
Fault detection and diagnosis.................................................................................................................... 43
PART II............................................................................................................................................................ 44
CHAPTER 3. ICA AT FINNISH ASPS .......................................................................................................... 46
3.2 PLANT CONFIGURATIONS AND OPERATION ................................................................................................. 46
Energy and chemical consumptions........................................................................................................... 48
Wastewater quality ................................................................................................................................... 49
Employees ................................................................................................................................................ 49
Wastewater quality ................................................................................................................................... 50
3.3 STATUS OF ICA IN FINLAND ..................................................................................................................... 50
Sensors..................................................................................................................................................... 50
Controls ................................................................................................................................................... 51
Alarm ....................................................................................................................................................... 52
____________________________________________________________________________________________
Advanced operation and control methods of municipal 2

wastewater treatment processes in Finland
Modelling ................................................................................................................................................. 52
Attitude towards ICA ................................................................................................................................ 53
Considerations from the WWTP visits ....................................................................................................... 54
CHAPTER 4. CONCLUSIONS AND FUTURE RESEARCH NEEDS ......................................................... 56
REFERENCES ................................................................................................................................................ 59
APPENDIX A: QUESTIONNAIRE ................................................................................................................ 67
APPENDIX B. KEY FIGURES OF THE WWTP SURVEY.......................................................................... 83
____________________________________________________________________________________________

Introduction
The importance of instrumentation, control and automation (ICA) at municipal wastewater
treatment plants (WWTPs) has increased as treatment requirements have tightened and the
processes have therefore become more complicated. A big effort has been done in recent years
in research and development work in this field: the reliability and accuracy of on-line sensors
has increased due to continuous work of manufacturers and researchers and in turn also the
applicability of sensor use in on-line control increased. Development of feasible soft-sensors for
different purposes in wastewater treatment has as well made operation of the plants more
advanced. In the past years computing power of the computers used to be a limiting factor in
use of ICA, but that is no longer case; practically free computing power is one of the most
important steps for possibilities of using advanced ICA not only at WWTPs but also in other
industrial works. Technical development has also been made regarding to different actuators
used in wastewater treatment which does enable more reliable and efficient control of the
processes.
In this Chapter the driving motivations for the present work are explained, as well as the
definition of the main goals are clarified.
1.1 Motivations
Since the implementation of the European Directive 91/271/CEE regarding urban wastewater
treatment, environmental water protection has gained increasing public awareness. Removal
requirements of ammonia have been changed to total nitrogen removal requirements during the
recent years also at many Finnish WWTPs and the trend is likely to continue. For this reason,
the number of unit processes has increased and many of them require specific conditions for
successful performance. Moreover, operation flexibility requires reliable on-line monitoring
and control methods for using the plant capacity efficiently. Plant complexity and the high
number of unit processes is one of the driving forces of increased need of ICA technology in
wastewater treatment. As different unit processes are based on chemical, mechanical and
biological phenomena, operation, control and monitoring of the whole plant are highly
challenging. In addition, A WWTP is characterized by frequent variations in environmental
conditions such as feed flow rate, temperature, influent concentration of nutrients and
concentration peaks of toxic substances that may cause serious problems in a biological
wastewater treatment. These variations can affect the process performance significantly,
sometime even resulting in process failures.
As number and reliability of on-line measurement sensors and analysers has increased one
important question arising considers possibilities of utilizing the large amount of process
information gathered. The process information should be possible to be easily refined to
practical reports for various purposes and target groups relating wastewater treatment plant
operation. Another highly important use for on-line measurements is their control applications.
Traditionally process monitoring has been the main purpose of major part of on-line
measurements; morever, in the near future also automatic fault diagnosis is likely to be a
substantial use for the sensor measurements.
Two main motivations in the implementation of ICA strategies for modern WWTP arise:
____________________________________________________________________________________________

Tighter effluent quality standards. To achieve the effluent quality set in the regulations
utilizing advanced ICA is beneficial; additionally, the effluent concentration can be kept
more stable and process faults disturbing the treatment can be decreased. As the number of
unit processes is increased becoming, for instance, also tertiary treatment processes more
common, the control of the whole plant more grows to be complex. The effluent quality can
be controlled efficiently by using modern ICA and achieving in this way even the tight
environmental regulations.
Cost Reduction. It is demonstrated (e.g. Olsson et al., 2005) that a good plant management
and ICA may increase the capacity of a nutrient removal WWTP by 10-30%, reducing the
necessarily need for enlargement of treatment basins when renovating the plant. Improving
the efficiency of the processes leads to smaller area needed for new WWTPs which means
in turns lower construction costs. In the next 10-20 years further understanding and use of
advanced ICA is predicted to reduce the total system investments by another 20-50%. In
addition, significant reduction in operation costs might be achieved in nutrient removal
plants relate to energy needed for aeration and various chemicals.
The wastewater treatment has become part of a production process where the quality control of
the effluent is very important since poor operation of the treatment process can lead to
important production losses and environmental problems, thus enhancement of optimal
operation and advanced control strategies have the potential to management of the plant in a
very successful way.
1.2 Objectives
For a long time, the application of process control and automation over a WWTP has been
encountering in the past years several difficulties, such as: (i) process complexity (including
chemical, mechanical and biological phenomena); (ii) difficulties in controlling the process
(due to the numerous parameters influencing the process for instance influent flow rate, organic
and nutrient load variations, toxicity, temperature, etc.); (iii) different dynamics for a process
that only seldom is steadystate conditions. The first need and goal of the present work has
been the understanding of the current direction in research and development of optimal
practices and advanced process control operation, by means of an exhaustive literature review
in this field. Taking as starting point the extensive technical report by Olsson et al. (2005), a
detailed information on the research projects executed during recent years is provided in the
publications referred in the literature review.
The background information on sensors, modelling and control introduces the most important
advanced control methods used in wastewater treatment worldwide. This gives useful cue to the
plant operators and engineers, a comprehensive overview on control solutions that could be
successfully implemented at WWTPs instead of the traditional options. In modelling sector as
well technical development has been made in recent years and some possibilities of dynamic
modelling and simulation in today and the trends of the future are introduced in this Report.
State-of-the-art surveys on ICA at WWTPs have been performed over the years with the
perspective of different countries. With this regard, the objective here is to assess and document
the current status of ICA in municipal WWTPs in Finland in order to determine successful
practices and the needs of plant operators. This gives also the possibility to compare ICA
situation among Finnish plants and also to the practices in other countries. The background
____________________________________________________________________________________________

information of the plant configurations, operation and wastewater quality is useful for people
involved in various sectors of wastewater treatment industry.
Eventually, recognizing the needs of further research and development in field of ICA is an
important goal of the present work since the development of ICA in wastewater treatment in
Finland in the future will be practically related to the state of ICA presented in this survey.
1.3 Report overview

This Report deals with advanced operation and control methods of municipal wastewater
treatment processes in Finland. It contains two main parts:
1. The first part reports the survey on the current status of research. The different aspect of
modelling and simulation (Chapter 2), control strategies (Chapter 3) and data analysis
(Chapter 4), are defined and the recent overcomes in academic word are analysed.
2. The second part reports the survey done in the Finland. The method of investigation is
based on a questionnaire including key elements regarding plant design, operation and
utilization of ICA, and operators opinion which was sent to large (> 100 000 p.e.),
medium-sized (30 000 100 000 p.e.) and small WWTPs in Finland. On-site surveys of
some of the most representative WWTPs were made.
Concise versions of the results reported in this work have been submitted to international and
national conferences.
____________________________________________________________________________________________

Part I
Literature Review
____________________________________________________________________________________________

Conferences on ICA in wastewater treatment have been held since 1973. Then the ICA
conference took place in London under sponsorship of IAWPR, International Association on
Water Pollution Research, presently called International Water Association (IWA). Back then
the lack of adequate instrumentation for the conditions of wastewater treatment was the key
issue. As computing power of computers was low at the time, hardly any closed loop plant
control was performed. In 1974 according to a survey done at some fifty WWTPs 12% of plants
in USA had automatic on-line dissolved oxygen (DO) control implemented. IAWPR
conference on instrumentation and control was held in Stockholm in 1977; by then progress in
DO and pH sensors had improved significantly their use in on-line monitoring and control.
Dynamic models of unit processes were presented. Computer was also used for several control
tasks such as DO, return sludge flow rate and bypass control at a WWTP in Sweden. Water
pollution control acts and funding programs began to encourage instrumentation and
automation of large and medium-sized plants in the 1970s.
In 1985 DO and suspended solids (SS) sensors were used on a routine basis and a self-tuning
DO control in full-scale was reported for the first time. By then micro-computers were well
established at WWTPs. In 1990 computer and communication technology were significant
driving forces for more advanced control; also new sensor technology and the first versions
commercial interactive simulators were introduced. In 1993 on-line nutrient sensors were
available. The role of dynamic modelling and simulation software were recognized important in
developing control strategies. Also new control methods such as neural networks and fuzzy
control became more common. In 1997 significant progress in models for on-line control had
been achieved. During 2000s there has been development in on-line nutrient sensors and
analysers, which are more robust and reliable enabling use of them in on-line control. Also new
optical DO sensor types have been introduced and become more popular in wastewater
treatment.
Olsson et al. (2005) provide a detailed history of ICA in wastewater treatment and of the related
conferences as well.
____________________________________________________________________________________________

Chapter 2. Modelling and Simulation
Traditionally, the formal modelling of systems has been done via a mathematical model, which
attempts to find analytical solutions enabling the prediction of the behaviour of the system from
a set of parameters and initial conditions. Modelling techniques include statistical methods,
computer simulation, system identification, and sensitivity analysis; however, one of these is as
important as the ability to understand the underlying dynamics of a complex system. Models
applied for prediction aim at providing an accurate and fast image of a real systems behaviour
under different conditions.
Models may be linear with respect to variables or parameters; furthermore, a model can be
nonlinear to parameters and linear to variables or vice versa. Linear models are used frequently,
because the analytical solution can be found. For nonlinear models numerical solutions are
predominant. Term mechanistic, physiological and white-box are used to describe that models
structure is based on physical, chemical and biological laws. Phenomenological, black-box,
empirical and heuristic the terms used for models that are based on empiricism rather than laws.
A combination of mechanical and phenomenological approach is called a grey-box modelling.
Validated process models can be used for dynamic simulations e.g. with different kinds of input
data. In simulations the mathematical equations of process model are solved and the results
given. Simulation is a reasonable way to extrapolate performance and scaling up process;
additionally, it helps in understanding behaviour and mechanisms of processes. Also the effects
of system parameters and disturbances can be investigated using process simulation.
The aim of this chapter is to review the current status of modelling and simulation of
wastewater treatment process.
2.1 Activated sludge process models

The activated sludge process (ASP) is the most widely used biological treatment of liquid
waste, essentially because it is a cheap technology which can be adapted to any kind of
wastewater. In the activated sludge process, a bacterial biomass suspension (the activated
sludge) is responsible for the removal of pollutants. Depending on the design and the specific
application, an activated sludge wastewater treatment plant can achieve biological nitrogen
removal and biological phosphorus removal, plus the removal of organic carbon substances.
Many different activated sludge process configurations have evolved during the years:
Jeppsson (1996) provides an exhaustive review on the historical evolution of the activated
sludge process.
Bioreactor models
There are several models describing the biological process in the activated sludge plant, the
developments in the family proposed by the International Water Association (IWA) represent a
major contribute. The models of ASM family (ASM1, ASM2, ASM2d, ASM3) are used in
most of the modelling and simulation studies (Henze et al., 2002); additionally, those are
considered state-of-art models of activated sludge processes and are used in most of the
commercial simulation platforms.
____________________________________________________________________________________________

When creating Activated Sludge Model No. 1 (ASM1) the aim for the IWA Task Group on
Mathematical Modelling for Design and Operation of Biological Wastewater Treatment was to
create a common platform that could be used for future development of models for nitrogen
removal activated sludge processes; furthermore, the aim was to develop a model with a
minimum of complexity. ASM1 was published in its final form in 1987. In ASM1 two kinds of
substrate, readily and slowly biodegradable COD (RBCOD and SBCOD), are introduced and a
hydrolysis process is included. In ASM1 it is also assumed that slowly biodegradable substrate
consists fully of particulate substrate (XS). ASM1 includes nitrogen and organic matter removal
with simultaneous consumption of oxygen and nitrate as electron acceptors; however, it does
not contain biological phosphorous removal. ASM1 was developed mainly for municipal
activated sludge plants (Henze et al., 2002; Henze et al., 2008).
Activated Sludge Model No. 2 (ASM2) was published in 1995; additionally, the model
included both nitrogen removal and biological phosphorus removal. The role of denitrification
in relation to biological phosphorus removal was still unclear, and Task Group decided not to
include that element. However, the development in research was fast, and denitrifying PAOs
(phosphorus accumulating organisms) were needed for simulation of many results from
research and practice. Because of this, the ASM2 model was expanded in 1999 into the ASM2d
model, where denitrifying PAOs were included (Henze et al., 2002).
Activated Sludge Model No. 3 (ASM3) describes the same processes as ASM1; however,
ASM3 was introduced to correct the deficiencies of ASM1. The most important reason for
introducing ASM3 was the recognition of importance of three rates of oxygen consumption in
the process: the rapid rate of oxygen consumption for degradation of RBCOD, slow rate
associated with degradation of SBCOD, and even slower endogenous oxygen uptake rate
(OUR). In ASM1 there is only one oxygen consuming process, which makes calibration of the
model very difficult. Calibration of ASM3 should be easier mainly because of converting the
circular growth-death-growth (death regeneration) model by endogenous respiration model
(Henze et al., 2008).
Secondary settler models

Activated sludge plants transform organic matter into biomass. The effective operation of the
process requires the biomass to be removed from the liquid stream (in the secondary settler)
prior to being discharged in the receiving waters. The sedimentation of the particles in the
liquor is achieved by gravity along with the density differences between the particles and the
liquid. Part of the biomass is purged, while a large fraction is returned to the biological reactor
to maintain the appropriate substrate-to-biomass ratio. This means that the settler combines
functions of clarification and thickening into one unit.
The complex behaviour of the secondary settler and its importance for the successful operation
of the ASP have made the settling process a great challenge for researchers working in the field
of mathematical modelling. A first physical model of batch sedimentation was developed by
Kynch (1952), in which the sludge transport is described by a mass balance partial differential
equation and that can be considered the origin of the solid flux theory. Different authors
introduced later several model extensions and among the others Petty (1975) considered the
continuous thickening of a solid-liquid suspension by gravity sedimentation; Vitasovic (1991),
Takcs et al. (1991) and lately Diehl (2007) considered the sedimentation flux from layer to
____________________________________________________________________________________________

layer. Jeppsson (1996) provides an extensive review on the different modelling approaches,
whereas to investigate the functioning of the settler, the IWA report (Ekama et al., 1997) gives
a complete reference.
In todays practice, Takcs model (Takcs et al., 1996) is by far the most widely used
mathematical representation of the secondary settler in published studies and commercial
software environments. However, some studies (as for instance Jeppsson and Diehl (1996),
Queinnec et al. (2001), Verdickt and Van Impe (2002)) stress the fact that the number of
discretization layers is used as a model parameter in order to match the experimental
observations. Typical number of 10 layers is used to introduce (artificial) numerical diffusion
and smooth off the concentration profiles. This model is therefore used without selecting a
number of layers in agreement with numerical convergence (the number of layers should be
selected large enough so that the numerical solution to the mass balance equations is computed
with an acceptable accuracy) and without distinguishing model formulation (i.e. the physical
model parameters) and numerical solution (i.e. the number of layers or grid points in a
numerical algorithm). Currently, David et al. (2009) consider the Method of Lines, which is a
straightforward two-step procedure, where the PDEs (Partial Differential Equations) are first
discretized in space, then integrated in time, can be used to solve the convectiondiffusion PDE
describing material transport in the secondary settler.
ASP Simplified models

Dynamic simulations based on rigorous and detailed modelling have become a standard tool in
many engineering fields. Rigorous models are applied for a variety of tasks: they allow the
exploration of the impact of changing some design configurations, can be used to provide tool
to actively explore new ideas and improve the learning process as well as allowing the
operators training facilities and thereby increasing their ability to handle unforeseen situations.
In other words, they are useful in understanding the system behaviour. Although the ASM
family models comprise much of the knowledge of the biological reactions a number of
drawbacks exist (Jeppsson, 1996): lack of identifiability, awkward nonlinearities, difficult
estimation and updating of time varying parameters. So, despite the usefulness of a rigorous
model, there are some situations where simpler models are better suited.
Different applications involves different requirement in the reduced model. For this reason
many reduced models for the activated sludge process are present in literature. In this section a
brief overview of some of the available models is presented, and for each case we discuss the
treatment goal for model reduction. In 1989, Marsili-Libelli developed a low order model for
conventional activated sludge systems with BOD removal and nitrification. The model was
developed to describe biodegradation of carbonaceous COD, nitrification, DO utilization, and
sludge sedimentation. As for the bacterial growth, a predator/prey modified Volterra approach
was used instead of the usually applied Monod model. The most straightforward model
complexity reduction is obtained by assuming only oxic conditions and thus neglecting the
denitrification processes. Kabouris and Georgakakos (1992) investigated the application of an
optimal control method to a reduced form of the ASM1 model obtained in such a way. For
Cartersen et al. (1995) the problem to be solved was an identification problem. They
formulated simplified models capable to give on-line information on the present state of the
wastewater treatment plant. The resulting model is a grey box model, where the Monod-kinetic
parameters of the nitrification and denitrification process can actually be identified and
estimated by means of prediction error decomposition and maximum likelihood estimation.
____________________________________________________________________________________________

They showed for the BioDeniPho system that it is possible to find a unique solution for each of
the unknown parameters of the model, starting from data collected in a real process. In 1995,
Zhao et al. introduced a simplified ASM1 model to describe the nitrogen dynamics in an
alternating activated sludge process. The proposed model is represented by a set of dynamics
equations in terms of ammonia and nitrate concentrations (on-line measurable). Isaacs (1996)
tested three model-based control strategies on a BioDeniPho system. All controllers employ a
relational model and a predictive model and different models were applied for different control
strategies. Along the same line Julien et al. (in 1998 and in 1999) developed a reduced order
model for identification and control of a single activated-sludge reactor operating nitrification
and denitrification. The reduced model is split into two submodels, one three-dimensional state
submodel in aerobic conditions (considering as state variables nitrate, ammonia and dissolved
oxygen) and one two-dimensional submodel in anoxic conditions, where only nitrate and
ammonia state variables are considered. The identifiability is based on on-line oxygen and
nitrate concentrations data, showing that the reduced model is structurally identifiable. Lately,
Gomez et al. (2000), Chachuat et al. (2003), Smets et al. (2003), Sperandio and Queinnec
(2004) and Mulas et al. (2007) proposed different approach to reduce the model complexity of
the ASM1.
An activated sludge plant in Avedre, Denmark, was modelled using Matlab/Simulink and
control strategies were tested with the calibrated model. A modified activated sludge model No.
2d (ASM2d) was used for modelling the activated sludge tanks and a reactive settler model for
modelling the secondary clarifiers. Also on-line recordings of all the direct and indirect control
handles were used as model inputs for calibration purposes. 30 days of control inputs preceding
the measurement campaign were applied to obtain proper initial conditions for the plant
simulation states (Ingildsen et al., 2006).
Before modelling a preliminary sampling campaign on influent was conducted to investigate

daily variations in the influent concentrations. After that a 13-day long measurement campaign
was done; furthermore, daily flow-proportional samples of the pre-settled influent and the
effluent were collected and analysed for different components. The proper model calibration
was done step-wise by distinguishing three subsystems: (1) the suspended solids system
(calibration of the sludge production and balances); (2) the nitrogen system (calibration of
nitrification and denitrification processes); and (3) the phosphorous system (calibration of P
removal processes). On-line sensor measurements as well as laboratory measurements were
used in the calibration the three subsystems. For testing of control strategies with model of
Avedre WWTP a benchmark system with a normal dry weather influent file and various
disturbance tests were prepared. The standard control strategy used at the WWTP was
simulated first for a validation. Variation for influent concentrations and flow rates were
created using a random number generator; thus, 90 days of variable influent concentration data
was created. In a successful phosphorous removal strategy also PHA (polyhydroxyalkanoate)
content is taken into account; moreover, that was noticed to reduce need of precipitation
chemical considerably (Ingildsen et al., 2006).
Three consulting teams conducted independent modelling projects at three different WWTPs in
order to evaluate and optimise the plant operation; two in USA (Denver, 1,800,000 p.e. and
Upper Marlboro, 250,000 p.e.) and one in Finland (Espoo, 250,000 p.e.) (Phillips et al., 2009).
The calibration methods and length of the sampling periods were different for each modelling
case. GPS-X simulation software was used in each of the simulation cases; with different
____________________________________________________________________________________________

activated sludge and clarifier models. Three common factors were found to be essential to
success of the projects: (1) additional sampling to define dynamic model inputs and influent
fractions; (2) careful planning; and (3) effective and frequent communication between all the
team members including the plant personnel.
In Espoo plant modelling case the most challenging process to calibrate was the primary
clarifier because models assume equal settling velocities for all particular COD fractions,
contrary to practical observations. In modelling the Espoo WWTP it was found that external
carbon was not optimal to dose in the last anoxic zone in order to avoid unnecessary carryover
of readily bio-available COD to the aerobic zones; in addition, the wastage of methanol could
be cut by 30%. It was also conducted that the yearly average nitrogen target of 70% could be
more cost-effectively reached by running the plant at 80% nitrogen removal in the summer and
60% nitrogen removal in the winter. Solid balance simulations showed that insufficient
thickening and digestion capacity and the overflow from the storage tank before centrifuges
were the main causes of problems with solids balance control. In Denver plant modelling case
the return sludge flow was optimized; likewise, the sizes of both anaerobic and aerobic zones
were optimized, and the risk of nitrifier washout at different design sludge retention times
(SRT) evaluated. In Upper Marlboro modelling case it was showed that maintaining the
existing three-sludge system provides for the most robust nutrient removal process of the
alternatives evaluated. (Phillips et al., 2009)
Rodrguez et al. (2009) presented a wastewater treatment modelling methodology based on

Excel and Matlab-Simulink for researchers without programming expertise. The proposed
framework also provides programming expert researchers a highly flexible and modifiable
platform on which to base more complex process model implementations. The use of Excel
interface for most of the implementation tasks makes the methodology accessible for non
programming researchers by providing them with clear overview of the models in a spreadsheet
file; in the same way, it shortens time and effort necessary to implement and modify model
structures also for researchers with programming expertise. Most of the information defining
the model, its parameters and a feeding schedule are provided using just an Excel interface.
Only the equations to calculate the rates of reactions and transfer processes have to be defined
in a Matlab file. The model implementation methodology presented has been successfully used
for a number of model implementations in research applications. It also provides a useful tool
for teaching by demonstrating the dynamics and factors affecting a wastewater treatment
process.
Modelling and simulation of wastewater treatment can also be combined with other software.
An interactive multiobjective optimization tool for decision support regarding to WWTP design
was proposed by Hakanen et al. (2008). Moreover, in the tool GPS-X simulation platform and
IND-NIMBUS method for solving industrial multiobjective optimization problems were
combined. The optimization problems produced by NIMBUS method are solved by using the
Controlled Random Search algorithm. In the case study presented a process model of nitrifying
activated sludge plant; in addition, ASM3 and the Takacs models were used in modelling the
activated sludge process and secondary clarifier. Residual ammonia nitrogen concentration,
alkalinity chemical dosing rate and aeration energy consumption were considered in the
optimization case task. In conclusion, the presented method was noticed to be promising and to
provide a good basis for further research including e.g. more complex processes, factors
affecting investment costs and more operational variables such as other chemical doses.
____________________________________________________________________________________________

2.2 IWA/COST Benchmark Simulation Models
A working group of COST Actions 624 and 682 together with the IWA Task Group for
Respirometry has developed a benchmarking protocol and software tool for benchmarking, i.e.
performance assessment and evaluation of control strategies for wastewater treatment systems
(Copp, 2002; Alex et al., 2008). The benchmark is a simulation environment defining a plant
layout, a simulation model, influent loads, test procedures and evaluation criteria. IWA/COST
Benchmark Simulation Model has recently been extended to include a complete WWTP (i.e.
wastewater and sludge treatment processes as well as interactions between them) (BSM2,
Jeppson et al., 2007). A simplified phenomenological model of a sewer system is also being
added (Olsson and Jeppson, 2006).
Based on the IWA/COST Benchmark Simulation Model No.1 (BSM1), the considered layout is
reported in Figure 1, it represents a fully defined protocol that characterizes the process
including a plant layout and two conventional control systems (Copp, 2002). The bioreactor
consists of five reactors, first two anoxic zones (pre-nitrification) followed by three aerobic
ones (nitrification). To maintain the microbiological population, the sludge from the settler is
re-circulated into the anoxic basin (returned activated sludge), and part of the mixed liquor is
recycled to the inlet of the bioreactor (internal recycle). The sludge concentration is kept
constant by means of sludge withdrawn pumped continuously from the settler. As for the
control, a DO controller is implemented to maintain the target aeration in the bioreactor;
besides, the DO level in the fifth reactor is controlled with a PI controller that manipulates the
aeration coefficient for this reactor, K5La. In addition, one control loop is used for controlling
the nitrate removal by manipulating the internal recycle flow-rate.
Figure 1. Plant layout for BSM1 Benchmark Simulation Model (Alex et al., 2008).
The plant is designed for an average influent dry-weather flow rate of 18 446 m3/d and an
average biodegradable COD in the influent of 300 g/m3. The influent dynamics are defined by
means of three files: dry weather, rain weather and storm weather; additionally, each of the
influent files consists of data of two weeks. The biomass sludge age used in BSM1 model is
about nine days (Alex et al., 2008).
The success of BSM1 as efficient platform for control strategies comparison in biological
nitrogen removal activated sludge plants is proven by the large number of scientific papers,
____________________________________________________________________________________________

already in 2004 more than 100 according to Jeppsson and Pons (2004), using the benchmark or
part of the benchmark (e.g. influent files, plant performance evaluation criteria).
The more recent BSM2 model (Jeppsson et al., 2004; Jeppsson et al., 2007) concerns the whole
plant including also primary clarifier and sludge treatment with anaerobic digestion. BSM2
implementations are available for wide range of simulation platforms including the most
popular commercial software. Even though the plant consists of several unit processes,
applying a supervisory control system taking into account the interactions between them is
beneficial for optimization of the plant operation. Furthermore, BSM2 aims at describing an
activated sludge plant with an influent load of 100 000 p.e. (80 000 from households and
20 000 from industrial origin) and it includes many of the main processes often used at large-
scale WWTPs (Figure 2).
Figure 2. Plant layout for BSM2 Benchmark Simulation Model (Jeppson et al., 2007).
The intensive use of BSM1 and BSM2 has also revealed a number of limitations, such as:
Too short evaluation period;
The seasonal effects are not taken into account;
Sensor, actuator and process faults are not included.
These represent the main reasons for the natural follow-up of the first benchmark with the
definitions of the long-term Benchmark Simulation Model No. 1 (BSM1_LT). Influent related
disturbances for BSM1_LT/BSM2 are generated with a model, and typical influent dynamic
phenomena generated with the BSM1_LT/BSM2 influent disturbance model, including diurnal,
weekend, seasonal and holiday effects, as well as rainfall, are illustrated with simulation results
in Gernaey et al. (2006). As a result of their work, the seasonal effects in terms of temperature
variations and changing influent flow rate patters are included in the evaluation period of one
year. The average temperature is 15C and the amplitude of sin wave shaped variation is 5C.
In addition, more than 60 control handles are available for the model.
____________________________________________________________________________________________

2.3 Plant-wide models
The second Benchmark Simulation Model BSM2 briefly introduced in the previous section
represents the extension of the BSM1 to more plant-wide point of view and actually also
different authors have faced plant-wide modelling for the entire WWTP taking into account the
dynamic description of the most relevant process in the water and sludge lines and the flows
among the different lines.
In particular, three different approaches for plant-wide modelling are presented and analyzed by
Grau et al. (2009): (i) direct connection among standard unit-process models (Interfaces
approach), (ii) modelling of the whole plant based on a common components vectors using
standard biochemical models (Standard Supermodel approach), and (iii) modelling of the
whole plant based on a common components vectors constructing biochemical models adapted
to the plant under study (Tailored Supermodelapproach). In the interfaces approach common
models of unit processes (e.g. ASM model family) are used. Main advantages of the approach
are the use of all knowledge previously acquired about the practical implementation of well-
known standard models and their simplicity; however, the main disadvantage is that all the
model components (e.g. COD fractions, the description of organic nitrogen, the description of
pH versus alkalinity) are not the same or even compatible. Therefore, general methodology for
connecting any two standard models has been proposed (Continuity-Based Interfacing
Methodology, CBIM). Elemental mass and charge continuity is guaranteed by means of a set of
instantaneous conversions from the original model components to the destination model ones
by a general procedure between any two standard models. Standard and Tailored Supermodel
approaches are based on a unique model for describing the most relevant biochemical, chemical
and physico-chemical processes in all unit processes of the plant. All the transformations are
therefore active in all of the streams regardless of the type of the stream or unit process.
Standard Supermodel approach is based on a set of standard models that describe the most
relevant processes within the whole WWTP. Supermodels (e.g. BNRM1, ASMD) reproduce the
activated sludge units for biological carbon, nitrogen and phosphorous removal, anaerobic
digestion, some chemical precipitation processes and additional calculations (e.g. DO, pH, pH
inhibitions) that can be switched on or off depending on the case study. The main advantage of
the approach is that the need for model interfaces is eliminated and any applicable
transformations are simply turned on or off. In contrast, the weak points of the approach are
lack of flexibility for describing simpler or more complex processes. The use of supermodels is
restricted to specific simulation platforms. The construction of a Tailored Supermodel is based
on the user selecting the set of compatible transformations strictly required to reproduce the
activity of the relevant bacteria populations in the WWTP under study. The main advantage of
the approach is the flexibility it includes to construct supermodels that are specifically adapted
to the requirements of the WWTP studied. However, the weak point is that the flexibility can be
easily misused if there is no rigorous and systematic procedure available to select the
transformations and to construct an appropriate model for each specific case.
A typical plant-wide model couples an upstream activated sludge plant (including primary and
secondary clarifiers) to an anaerobic sludge digester. One of the key challenges has been the
definition of an interface between activated sludge model (ASM1) and anaerobic digestion
model (ADM1) the state variables of which differ to each other. A new interface and
characterization model (the modified Copp interface) was proposed by Nopens et al. (2009);
furthermore, the approach was demonstrated both hypothetically (BSM2 Benchmark
Simulation Model) and practically on a full-scale WWTP. ADM1 model has, as inputs, a
____________________________________________________________________________________________

mixture of carbohydrates, proteins, lipids, organic acids, bicarbonate and ammonium, some of
which are difficult or even impossible to measure; previously this has been tried to overcome
using a generalized pool of complex material X C (the Copp interface). In the modified Copp
interface approach the degradable components are converted directly to sludge fractions used in
ADM1 model instead of using XC. The steps of converting the state variables from ASM1 to
ADM1 are presented in Figure 3. The behaviour of the modified Copp interface was tested
exactly the same way as the Copp interface using Benchmark Simulation Model BSM2;
additionally, improved degree of realism was noticed regarding to the composition of anaerobic
digester effluents and gas production. To validate the proposed model a practical test case using
data of a full-scale digester in Brisbane, Australia, was executed. The feed of the digester
consisted of a mixture of primary and secondary sludge. The model predictions were in good
agreement with the dynamic measurements of gas flow rates. Moreover, the new interface
allows the use of one interface for a range of input sludge resulting in realistic gas production
without need for extensive parameter calibration.
Figure 3. Steps of converting state variables from ASM1 to ADM1 in the modified Copp
interface approach (Nopens et al., 2009).
2.4 Process Simulation Environment

The knowledge acquired about the process in a wastewater treatment plant can now be used to
implement appropriate dynamical simulations of it. By using simulations, we can study the
effect of different environmental conditions, test the system sensitivity to different parameters
and apply different control configurations and so on. Information about specific WWTP
simulators is given by Olsson and Newell (2001) and also by Copp (2002). From a practical
stand point, a simulation can be obtained in different ways either using commercial software or
implementing the ASP model in programming environment like Matlab/Simulink and/or
Fortran. In the following section a description of the two approaches is given.
Specific commercial environments usually contain extended libraries of predefined process

models offering the representation of the whole wastewater treatment plant. The process
configuration to be simulated can be easily constructed by connecting process unit blocks and
pop-up windows allow modifying the model parameters.
____________________________________________________________________________________________

Many examples of commercial simulators exist. Among them we can list the following:
EFOR by Danish Hydraulic Institute (Denmark) is a software tool dedicated to the
modelling of wastewater treatment systems, which makes it easy to construct a wide variety
of wastewater treatment plants. The present progress can be found on
http://www.dhisoftware.com/efor/. EFOR 3.0 includes ASM1 and ASM2 model, which is
modified to include biological phosphorous removal; furthermore, three different hydraulic
models are available for the settlers. Treatment plants with a complex and dynamic
operation can be described and simulated with EFOR. The software is available in
Windows environment.
BioWin by EnviroSim Associates Ltd (Canada) is a dedicated process simulator that makes
use of linked process units to simulate biological wastewater treatment systems. BioWin is
a Microsoft Windows -based simulator used in the analysis and design of wastewater
treatment plants. Engineering companies are the most important user group of BioWin and
most of users are located in Northern America. BioWin Version 3.0 contains an integrated
biological model for biological nutrient removal (BNR) activated sludge, fermenters,
Moving Bed Biofilm Reactor (MBBR) & Integrated Fixed Film Activated Sludge (IFAS)
systems, and anaerobic digesters. The model utilizes BioWins full General Activated
Sludge/Anaerobic Digestion Model (ASDM) that tracks over 50 components with more
than 80 processes acting on these components. More information can be found at
http://www.envirosim.com/products/bw32/bw32intro.php.
GPS-X by Hydromantis Inc, Canada (http://www.hydromantis.com/), is a modular,
multipurpose modelling environment for the simulation of wastewater treatment systems.
GPS-S Version 5.0 is supplied with over 50 preconfigured layouts covering most of the unit
processes found in wastewater treatment plants. The simulator is built on the ACSL
simulator, that provides powerful integration and general simulator features. Six standard
biological models e.g. temperature dependant versions of ASM1, ASM2d and ASM3 are
available in GPS-X. The biological unit processes include carbon, nitrogen and phosphorus
removal, in various suspended growth and fixed film configurations.
SIMBA (SIMulation programms fr die Biologische Abwasserreinigung): developed at the
Institut fr Automation und Kommonikation (IFAK) in Germany. It can be considered a
custom made version of Simulink for wastewater treatment applications; with its latest
version SIMBA 5, allows the holistic consideration of sewer system, wastewater treatment
plant, sludge treatment and rivers. It extends Matlab/Simulink using block libraries for
biological and chemical treatment processes. SIMBA includes several default models
including ASM1, ASM2d, ASM3, the Bio-P Model and several settler models. A more
extensively compendium about the simulator can be found on http://simba.ifak-
md.de/simba/.
STOAT (Sewage Treatment Optimization and Analysis over Time) by Water Research
Center (UK) is a Windows based computer-modelling tool designed to dynamically
simulate the performance of a wastewater treatment works including sludge treatment
processes. STOAT includes an implementation of ASM1, called IAWQ No.1, and the
Takcs settler model, called Generic. The software can be used together with commercial
sewerage and river quality models. Further information about STOAT software is on the
following website: http://www.enviatec.de/en/en_start_stoat.htm
WEST (Wastewater treatment plant Engines for Simulation and Training): an interactive
dynamic simulator. It is developed mainly at the University of Gent, Belgium and current
information about the software can be found on http://www.hemmis.com/. WEST includes
a number of modules and features that enables the user to model and evaluate almost any
____________________________________________________________________________________________

kind of wastewater treatment plant application that exists. Most of the models in WEST
simulator are open source and open code; thus, the models can be modified if necessary.
WEST mainly has been used in the context of wastewater treatment research.
Matlab is a general high-level language for technical computing. It includes a large library of
predefined mathematical functions. Additionally, it features a family of specific toolboxes that
extent the Matlab environment to solve particular classes of problem (there are approximately
40 toolboxes available). Moreover, Simulink is an add-on software product to Matlab for
modelling, simulating and analyzing any type of dynamic system. Matlab and Simulink are
fully integrated, meaning that all functionalities of the Matlab toolboxes are available in the
Simulink environment as well. Simulink provides a graphical user interface for building models
as block diagrams and manipulating these blocks dynamically. A large number of predefined
building blocks are included and it is easy to extend the functionality by customizing blocks or
creating new ones. The capabilities of Simulink may be further extended by using the S-
functions (system functions), which can be written in Matlab language, C++ or Fortran using
predefined syntax. Consequently, S-functions can be easily incorporated and a dynamical
system can be described as a mathematical set of equations instead of using predefined block
diagrams.
____________________________________________________________________________________________

Chapter 3. Process control
In modern wastewater treatment plants processes such as aeration, chemical feeds and sludge
pumping are usually controlled by on-line sensor measurements. Different control algorithms
are used in manipulating actuators such as valves and pumps in order to keep process
conditions favourable for demanded treatment results and cost-effective. For example in
aeration zones there is set-points for dissolved oxygen (DO) concentrations and aeration is
controlled according to difference between DO set-point and measurement; furthermore,
control algorithms are used in adjusting aeration valve positions and air compressor operation.
Conventional and some of the advanced control algorithms used in wastewater treatment are
presented in this chapter.
3.1 Feedback control algorithms

Typical components of a feedback control loop are presented in Figure 4. Overall control
system performance depends on proper choice of each components of a feedback control loop.
Figure 4. Typical components of a feedback control loop.

Two types of algorithms predominate in WWTPs, and in the process industry in general, the
on-off and the Proportional-Integral-Derivative (PID) algorithms. In the following of this
section, both of them will be discussed.
On-off control
On-off controllers are simple, inexpensive feedback control in which the controller switches the
actuator between two stages according to sensor measurements and the control law. Thus, the
controlled variable is kept within certain limits. In on/off controller the control variable can
assume only two values, umax and umin, depending on the control error (e) sign. The control law
is defined as follows:

umax if e > 0
u=
umin if e < 0

When using on/off controller, the process variable oscillates persistently around the set-point,
which can be noticed in the example of Figure 5; additionally, in the example umax of the
controller is 2 and umin is 0. On-off controller is often modified by inserting dead zone (resulting
in a three-stage controller) or hysteresis in order to cope with measurement noise and to limit
the wear of the actuating device. Hysteresis is a phenomenon, which causes a time lag in
____________________________________________________________________________________________

application of a control action (Visioli, 2006). In spite of that, the main disadvantage of the on-
off control is that it results in continuous cycling of the controlled variable and produces
excessive wear on the final control element.
Figure 5. Example of an on/off control application. Solid line: process variable; dashed line:
control variable (Visioli, 2006).
Turning the aeration system on and off as the oxygen level falls below or rises above a given
DO level in the aeration tank is an typical example of on-off control in wastewater treatment.
Another example is the control of the water level in the pumping station or in operating pumps
in return and excessive sludge flow control loops (Tchobanoglous et al., 2003).
PID algorithms
PID is the most common control algorithm used in process industry and wastewater treatment.
The letters PID stand for proportional, integral and derivative. It is a control method in which
the controller output is proportional to the error (P), its time history (I), and the rate at which it
is changing (D). Basically, it is the combination of the three control actions, proportional,
integral and derivative; moreover, the choice of the used combination depends on the
application being the PI combination the most common. PID controller has been in use in
pneumatic form, in the form of an analogue electronic circuit, as a digital circuit, and in
software using a microprocessor (Ellis, 2004).
P term is proportional to error (e) between measured value and set-point. Parameter KP is the
proportional gain and it determinates how strong the control action of the P term is. It is typical
that P term alone is not sufficient to compensate whole of the error variable. In the same way,
the integral term integrates error variable over time. The contribution from this term is
proportional to both the magnitude of the error and the duration of the error. Also, it accelerates
the movement of the process towards set-point and eliminates the residual steady-state error
that occurs with a proportional (P) only controller. Likewise, derivate term D examines the
change rate of the error variable. The D term tries to predict and compensate the offset at the
same time when it is setting, while parameter KD represents the derivate gain. In brief, the
proportional, integral, and derivative terms are summed to calculate the output of the PID
controller. A block diagram of a PID controller is presented in Figure 6.
____________________________________________________________________________________________

Figure 6. Block diagram of a PID controller (Visioli, 2006).
Defining u(t) as the controller output, the equation of the PID algorithm is:
t
de(t)
e(t)dt +
1
u(t) = u0 + K P e(t) +
I
D
0 dt
where:
u0 is the bias value;
KP is the proportional gain (usually dimensionless);
I is the integral time or reset time (it has units of time);
D is the derivative time (it has units of time).
The definition of the values for the parameters KP, I and D represents the tuning of the control
loop; many techniques have been proposed for this task and among the numerous books,
Seborg et al. (2003) represent a good reference for the basic theory in process control.
When choosing the controller for an application, complexity must be weighted against
performance. More complex controllers are also more difficult to tune; additionally,
considerable technical expertise is required to understand how advanced controllers work and
how to fix them when they fail. PID controllers are predominantly used to maintain the constant
value of a process variable under conditions of changing load. Further, for the majority of
process control tasks a PI controller is capable of providing an acceptable performance and it is
the most adopted controller in the industrial context (Visioli, 2006).
Although many advanced control algorithms have been proposed, conventional feedback PID
control algorithms are the most popular in wastewater treatment. Also in other typical process
control applications like pulp and paper industry more than 95% of controllers are PID
controllers (Henze et al., 2008).
3.2 Feedforward Control

With feedforward control the basic idea is to measure important load variables and take
corrective action before they upset the process. In contrast, a feedback controller does not take
corrective actions until after the disturbance has upset the process and generate an error signal.
The main idea of feedforward control consists of measuring disturbances and changing the
manipulated variables so that the output of the controller will be maintained at desired range.
For example, if the influent characteristics and flow rate (disturbances) are measured and it was
possible to calculate the required change in airflow (manipulated variable) supplied to an
activated sludge process to maintain constant the dissolved oxygen concentration in the aeration
tank (controlled variable), it would be possible to implement feedforward control. In practice,
____________________________________________________________________________________________

the dynamic model of the process must be known and its mathematical model has to be
available.
Most of the measurements of the disturbances as well as the mathematical model calculation
are very difficult tasks. As a result, feedforward control has limited application in WWTPs. The
current limited applications of feedforward control include control of chemical addition and
control of return activated sludge flow from the secondary settler to the aeration basins
(Tchobanoglous et al., 2003).
In practical application, feedforward control is normally used in combination with feedback

control; this combination can provide a more responsive, stable and reliable control system.
Combined feedforward and PI control has been proposed e.g. for external carbon flow control
(Samuelsson and Carlsson, 2001) and DO concentration control (Yong et al., 2005) in activated
sludge processes.
3.3 Advanced control algorithms

As processing plants become increasingly complex in order to increase efficiency and reduce
costs, there might be greater incentives for using advanced control. This section will emphasize
on the techniques encountered in WWTP applications.
Cascade Control
A disadvantage of feedback control is that corrective action for disturbances does not begin
until after the controlled variable deviates from the set-point. As discussed in the previous
section, feedforward control requires that disturbances be measured explicitly and that the
model be available to calculate the control output. An alternative approach, that can
significantly improve the dynamic response to disturbance, employs a secondary measurement
points and a secondary feedback control. The secondary measurement point is located so that it
recognises the upset condition sooner than the controlled variable, but disturbance is not
necessary measured. This approach is called cascade control: one feedback controller, identified
as the primary loop, is used to calculate the set-point of another feedback controller that
represents the secondary loop. Interaction among the control loops is the most pressing reason
for utilization of cascade control.
This approach is widely used in process industries and it is particularly useful when the
disturbances are associated with manipulated variables, in wastewater treatment plants is used
mainly for the following reasons (Olsson and Newell, 1999):
Disturbance rejection: DO controllers are often cascaded onto a flow or pressure slave
slope. This filters out the disturbances caused by changes in the air supply pressure before
the DO concentration is affected.
Gain scheduling: one or more observable variables, called the scheduling variables, are used
to determine what operating region the system is currently in and to enable the appropriate
linear controller.
Hysteresis removal: a secondary control loop can be added to supply the strong control
signal and remove the oscillations. Valve positioners on control are the most common
examples.
A further cascade control strategy is proposed Cho et al. (2002); the controller is composed of
two Proportional-Integral (PI) controllers to regulate the nitrate concentration in the pre-
____________________________________________________________________________________________

denitrifying process by manipulating the external carbon dosage as reported in Figure 7. It
controls the nitrate concentrations in the effluent as well as in the final anoxic reactor
simultaneously to strictly satisfy the quality of the effluent as well as to remove the effects of
disturbances more quickly. Assuming that the concentration of outflow of the fifth
compartment is the same as that in effluent due to no chemical reaction in the secondary settler.
Then, the primary loop is composed of the nitrate concentration in the fifth compartment as the
process output and the set-point of the secondary controller as the process input. The secondary
loop consists of the nitrate concentration in the anoxic zone as the process output and the flow
rate of the external carbon source as the process input. The proposed approach using only the
external carbon source dosage shows better regulation results. It is also advantageous in the
maintenance and economic aspects, using smaller number of sensors than feedforward control.
Figure 7. Configuration of cascade control loops in Cho et al. (2002).
Model predictive control

The basic concept of model predictive controllers (MPC) is to choose number of future
adjustments of the manipulated variables to minimize the errors between the controlled
variables and their set-points, based on the predicted response of the process. MPCs
traditionally use simple form of linear process models, but in recent years also nonlinear models
have become more common. MPC is capable of handling multivariable processes, which has
made it a popular control method. In addition, the prediction horizon keeps being shifted
forward and for this reason MPC is also called receding horizon control. There are several
commercial MPC software implementations, but they are relatively simple to implement for
anyone familiar with modern matrix manipulation techniques. Basic structure of MPC is
presented in Figure 8.
Figure 8. Basic structure of MPC (Camacho and Bordons, 2004).

____________________________________________________________________________________________

All MPC algorithms possess common elements, and different options can be chosen for each of
the elements. These elements are (Camacho and Bordons, 2004):
Prediction model;
Objective function;
Obtaining the control law.
The model is the core of MPC; furthermore, it should be complete enough to fully capture
process dynamics and allow the predictions to be calculated. The different strategies of MPC
can use various models to represent the relationship between the output and the measurable
inputs, some of which are measurable variables and others can be considered as measurable
disturbances which can be compensated for by feedforward action. Chosen model structure is
usually a compromise between process characteristics and control objectives. Developing a
functional model is the most critical and time-consuming part of designing MPC. When
applying MPC, it has to be ensured that the process model employed represents the processes in
sufficient accuracy. The model can be separated in two parts: the actual process model and the
disturbances model, which describes e.g. the effect of non-measurable inputs, noise and model
errors. (Posio, 2002; Schtze et al., 2002; Camacho and Bordons, 2004)
MPC approach to controller design involves on-line optimization calculations, which take
account of system dynamics, constraints and control objectives. Various MPC algorithms
propose different cost functions for obtaining the control law. In MPC control future control
actions are counted by minimizing objective function of performance over a certain prediction
horizon. Control calculation is carried out as a result of three steps:
1. Use of model in predicting the output variables of the controlled process as future discrete
over prediction horizon.
2. Calculation on future control sequence over control horizon by minimizing given objective
function within constraints in a way that the predicted process output variable is as close as
possible to desired reference signal.
3. In strategy of receding horizon only the first output value of control sequence received from
minimizing procedure of objective function is implemented in process control and horizons
are slid towards future. Optimization is repeated with updated data on the next sampling
time.
Steps 1 and 2 can be understood as control actions of open loop whereas repeating calculation
in step 3, implementing the first value of control sequence to system control and discarding rest
of the values closes the loop.
MPC is widely used in process industries in particular (Qin and Badgwell, 2003). In wastewater
treatment behaviour of the processes is often nonlinear, which makes control challenging. MPC
has been used at WWTPs e.g. in controlling the pH value (Camacho and Bordons, 2004) and
dissolved oxygen concentration (Cristea and Agachi, 2006).
Lately, Shen et al. (2008) applied MPC to the BSM1 wastewater treatment process to maintain
the effluent quality within regulations-specified limits. Good performance was achieved under
steady influent characteristics, especially concerning the nitrogen-related species. The results
showed that the efficiency of operating biological wastewater treatment processes could be
____________________________________________________________________________________________

significantly influenced by an overload in a local community due to varying wastewater
sources, chemical composition and flow rate. When the plant is highly loaded and if stringent
effluent fines are imposed by legislation, in coping with this problem and performing a cost
effective operation, the more advanced control algorithms, such as model predictive control
scheme, are advantageous.
Fuzzy logic control

Fuzzy control algorithms use fuzzy logic to allow machines to make decisions based on the
practical knowledge of a human operator. Fuzzy logic is a mathematical system that analyzes
analog input values in terms of logical variables that take on continuous values between 0 and
1, in contrast to classical or digital logic, which operates on discrete values of either 0 and 1
(true and false). Fuzzy control method can be implemented in similar control applications as
e.g. PID control and a model-based control system. Fuzzy control contains (nonlinear) mapping
that has been defined using fuzzy IF-THEN statements. Applying fuzzy logic calculations
enables to design a nonlinear controller, without a detailed knowledge of the operating point
nonlinearity, as would be required for a classical control design.
A fuzzy control device consists of three stages, all of which perform calculations: an input
stage, a processing stage and an output stage. The control system utilizes a database with
includes fuzzy rule base and the numerical information needed in calculation.
The control is made in terms of a rule base that performs operations on the fuzzy sets and
interference. In the input stage the input signals translated to fuzzy logic values (fuzzification).
Multi-valued logic inference rules are applied to develop the control action during the
processing stage. In the output stage the fuzzy logic values are re-translated to continuous
signals (defuzzification), which are used as control outputs. A generic fuzzy system is presented
in Figure 9 (Blevins et al., 2003).
Figure 9. A generic fuzzy system with fuzzification and defuzzification units and external
dynamic filters (Verbruggen and Babuka, 1999).
At first the fuzzy sets are defined and it is decided how to logically quantify control inputs. The
typical fuzzy sets for input signals are Negative Large, Negative Medium, Negative Small,
Zero, Positive Small, Positive Medium, and Positive Large.
____________________________________________________________________________________________

Decision making in fuzzy control follows IF-THEN rules, for example:
IF such and such states of process variables (i.e. inputs) are
THEN such and such control actions (i.e. outputs) are needed
To establish boundaries for the set, membership functions are defined analytically for every
fuzzy set. Membership functions can be of various types: linear, parabolic, exponential or
normal distribution. Linear membership functions are used most often due to the simplicity and
satisfactory flexibility. When developing fuzzy inference rules, it is important to understand
process behaviour. In fact, the most common shape of membership functions is triangular. Also
trapezoidal and bell curves are used (Figure 10). (Blevins et al., 2003)
Figure 10. Various types of membership functions representing fuzzy sets (Verbruggen and
Babuka, 1999).
There are many different methods of defuzzification in which the control output value is
determined. The centroid method, in which the "center of mass" of the result provides the crisp
value, is very popular. Another approach is the height method, which takes the value of the
biggest contributor (Kova and Bogdan, 2006).
Use of fuzzy controller can be reasonable for various purposes in wastewater treatment (1) if
there is no analytical model of the WWTP available, (2) if the model is too complicated to
compute a controller by classical methods or (3) if the control goals are not defined precisely.
Fuzzy logic control does also have some downsides, which are described in the following.
(1) Knowledge acquisition can be very troublesome. It is often hard even for an experienced
operator to express how control should work. (2) Tuning a fuzzy logic controller is difficult
because there are no systematic tuning methods for them. Often set of rules for the control has
to be tuned many times by trial and error before the demanded level of performance is
achieved. This can be troublesome and take lots of time. (3) Because fuzzy logic control does
not exactly model the controlled process, it is difficult to prove the control to be optimal and
stabile. The larger the set of rules is, the bigger the problem often is. (4) Testing the control is
difficult if the fuzzy set of rules is large. When the size of set of rules enlarges, becomes testing
more complicated because of interactions between the rules.
Fuzzy logic control has been applied e.g. for controlling of sludge height in secondary settler of
a WWTP (Traor et al., 2006) and in optimizing volume distribution in each stage of a step
feeding activated sludge process (Zhu et al., 2009). Fuzzy control has also been used in
controlling nitrate recirculation flow and external carbon addition in a pre-denitrification
____________________________________________________________________________________________

process using oxygen reduction potential (ORP) measurement as a control parameter (Peng et
al., 2005).
Artificial neural networks

Neural Network (NN) model is a complex interaction between fairly simple expressions called
neurons. NN models are able to represent both linear and nonlinear relationships. Artificial
Neural Network (ANN) is an information processing paradigm that is inspired by the way
biological nervous systems process information. ANN consists of group of artificial neurons.
Neurons are simple data processing units connected to each other. Furthermore, neuron consists
of the following units: (1) synapse, (2) adder and (3) activation function (Lingireddy and Brion,
2005).
Figure 11. Schematic representation of a typical ANN (Lingireddy and Brion, 2005).
ANN model is a simplified view of the structure of brain combined with mathematical logics.
The network is composed of a large number of highly interconnected processing elements
(neurones) working in parallel to solve a specific problem. ANNs are taught by examples,
which should be selected carefully, and they can not be programmed to perform a specific task.
In most cases an ANN is an adaptive system that changes its structure based on external or
internal information that flows through the network during the learning phase. Before learning
ANN typically starts out with randomized weights for all their neurons. The ability of ANNs to
approximate complex nonlinear relationships without prior knowledge of the model structure
makes them attractive alternative to classical modelling and control technology. ANN consists
of neurons divided to input layer, hidden layer(s) and output layer (Figure 11). The activity of
the input units represents the raw information that is fed into the network. In the input layer
there is one neuron for each model input. There are one or more hidden layers in ANN. The
hidden layers perform a weighted sum of inputs from each neuron of the previous layer,
transform the sum according to some activation function and distribute the result to each neuron
of the next layer. Subsequently, the output layer produces the final output. Each neuron takes
many input signals and based on an internal weighting system, produces a single output signal,
which is typically sent as input to another neuron (Lingireddy and Brion, 2005).
Radial Basis Function Networks (RBF) is type of ANNs, which have only one hidden layer.
RBF has been shown to be better than traditional ANN in fault detection applications. In RBF
inputs are process measurements and outputs each indicate a certain fault or operating condition
(Olsson and Newell, 1999).
____________________________________________________________________________________________

In the process industries, one frequent application of NNs is as a soft sensor. When many
physical measurements about the process are known, the value of an unmeasured variable can
be computed, in real time, using a neural network software program. In wastewater treatment
hybrid time-delay neural networks (TDNN) combined with principal component analysis
(PCA) has been used in a soft sensor predicting effluent BOD value (Zhao and Chai, 2005).
ANN has also been used in control of coagulation, flocculation and sedimentation processes in
water treatment (Lingireddy and Brion, 2005) and to keep the concentration of the recycled
sludge proportional to the influent flow rate in presence of disturbances, parameter variations
and measurement noise (Baruch et al., 2005). Neural networks have great potential in control of
wastewater treatment processes in general and anaerobic sludge digestion in particular (Olsson
et al., 2005).
3.4 Control applications

IWA/COST Benchmark Simulation Models have used in number of studies on different control
strategies during the recent years. Also other modelling and simulation environments have been
used in several of the research papers. Real plant experiments have been relative rare, and often
testing of control strategies using dynamic simulation have preceded real plant tests.
BSM1 applications
BSM1 protocol has been often used in studies on different control strategies, operational cost
estimation and risk evaluation in wastewater treatment.
Six different integrated control strategies in pre-denitrifying ASP were tested by Yong et al.
(2006) using the IWA/COST Benchmark Simulation Model BSM1. The integrated control
strategies concern nitrate recirculation flow rate and external carbon dosage. ASM1 model was
used for modelling the wastewater treatment process and the controllers were implemented
using Matlab/Simulink software. The proposed control strategies are presented in Figure 12.
The nitrate recirculation flow rates are increased at night (low-load periods) maximizing the use
of influent and intercellular COD for denitrification. During daytime (high-load periods) nitrate
recirculation flow rates are lower. Furthermore, external carbon dosage is low at night and high
at daytime. Compared to pre-defined simple control example all the integrated control strategies
were noticed to improve significantly the effluent quality; the average nitrate and total nitrogen
concentrations in the effluent were reduced by 42-47% and 29-33% respectively. However, the
effluent ammonium concentration was increased by 3.5-13%.
Integrated control strategies for nitrogen removal in activated sludge processes were studied
using BSM1 by Yong et al. (2006). The strategy no. 1 was concluded to be the best one for
nitrate recirculation flow rate and external carbon dosage in terms of external carbon
consumption and plant performance criteria. The strategy makes best use of plant denitrification
capacity and maximizes the use of influent COD for denitrification. In the control strategy no. 1
there are two feedback control loops; one to determinate the flow rate of external carbon source
to the first anoxic zone in order to keep the nitrate concentration at the defined level at the end
of the second anoxic zone, and the other one to adjust the flow rate of the nitrate recirculation to
maintain the nitrate level at defined level at the end of last aerobic zone.
____________________________________________________________________________________________

Figure 12. Six integrated control strategies of nitrate recirculation and external carbon addition
(Yong et al. 2006).
Samuelson et al. (2007) researched optimal set-points and cost minimizing control strategies for
denitrification process in the activated sludge process using BSM1; nitrification process was
not considered in the study. The manipulated variables are the internal recirculation flow rate
and the flow rate of an external carbon source; while, the controlled variables are the nitrate
concentrations in the last anoxic zone and the effluent. Three dynamic data input files are
defined in the Benchmark Simulation Model, each describing different weather conditions. The
stationary operational costs of the denitrification process were analysed and presented in
stationary operational maps together with considered output signals (Figure 13). The partial
costs taken into account were: (1) pumping costs (required energy), (2) aerations costs (required
energy), (3) external carbon source costs, and (4) possible fee for effluent nitrate discharge. The
simulations were done e.g. with different energy prices, with a cost-free external carbon source
and without a fee for nitrate effluent load. As a result, it was found out that cost-optimal set-
point of nitrate in the anoxic zone depends on the choice of effluent nitrate set-point.
____________________________________________________________________________________________

Figure 13. Stationary operational map for a grid of different values of external carbon source
flow (Qcar) and internal recirculation flow (Qi) for the case with no nitrate fee. Solid lines show
the total cost in EUR, dash-dotted lines show effluent nitrate concentration S eNO (mg(N)/l), and
dotted lines show the nitrate concentration in the last anoxic compartment SanNO (mg(N)/l)
(Samuelsson et al., 2007).
There are several possibilities how to design the actual control law in order to minimize
operational costs: (1) using two different control loops to control the effluent nitrate
concentration SeNO and the nitrate concentration in the last anoxic compartment SanNO
separately, (2) using constant high internal recirculation flow rate (since the impact of it on total
cost is much smaller than the impact of external carbon source flow), (3) using quadric criteria,
for example Linear Quadratic Controllers (LQC) or Model Predictive Controllers (MPC), on-
line for minimization of total cost, or (4) using a simple grid search on-line until the optimal
point is reached. The presented operational cost minimization tool could be valuable in running
WWTP in a more cost effective way, but before that the approach should be extended to also
consider nitrification process, and it should also be evaluated using on-line data from a full-
scale WWTP (Samuelson et al., 2007).
Stare et al. (2007) proposed several control strategies for nitrogen removal. The strategies were
evaluated using BSM1 and operating cost was used as evaluation criteria; moreover, the
operating cost consists of aeration costs, sludge disposal costs, external carbon dosage costs and
effluent fines. The manipulated variables used in the study were external carbon flow rate
(Qcar), DO set-point (Soset) and oxygen transfer rate (KLa) while the other variables were set to
the constant values. The five different control strategies studied were (a) constant manipulated
variables, (b) oxygen PI control, (c) nitrate and ammonia PI control, (d) nitrate PI and ammonia
feed forward-PI control, and (e) model predictive control (Figure 14). The more advanced the
control strategy is, the more on-line measurements and additional sensors are needed. In the
control strategy (e) a multivariable and nonlinear MPC was implemented.
____________________________________________________________________________________________

Figure 14. Control structures: (a) constant manipulated variables, (b) oxygen PI control, (c)
nitrate and ammonia PI control, (d) nitrate PI and ammonia FF-PI control, and (e) model
predictive control (Stare et al., 2007).
Optimal operating set-points points were determined using operational maps and each control
strategy was simulated using the optimal set-point values. The results showed that with nitrate
PI and ammonia FF-PI control almost the same operating cost can be achieved as with
advanced MPC algorithms. When the plant is highly loaded the MPC algorithm was noticed to
be advantageous. Thus, for sensitive discharge points where preventing high peak values in the
effluent is important implementing an advanced control algorithms is reasonable. For less
sensitive water bodies using simple control algorithms was concluded to probably be sufficient
(Stare et al., 2007).
Control of aeration volume in an ASP was investigated by Ekman et al. (2006). In the proposed
control strategy DO concentration in some of the compartments is determined by a higher level
controller driven by the DO concentration in other zones; thus, DO set-point is time-varying for
some of the zones. BSM1 benchmark protocol was used in the research; similarly, the strategies
were also tested by test runs at a pilot plant in Hammarby Sjstad (Stockholm). The main idea
of the suggested aeration volume control strategies can be described as follows: (1) during
periods of low influent load a relatively low number of compartments are aerated using
controllers with pre-specified DO set-points, and (2) when the influent load increases, some
additional compartments are also aerated. In the suggested cascade control strategy extra
supervisory (master) PI controllers are implemented; furthermore, the supervisory controllers
have to be tuned so that the inner control loop is much faster than the outer control loop. The
main idea for the cascade control of the first and the last aerobic compartments is to aerate the
compartment only when necessary. Simulation studies were conducted for the suggested
cascade control strategy, as well as for a constant DO control strategy and for a supervisory set-
point control strategy based on ammonium concentration measurements. In pilot plant test runs
it was concluded that the suggested aeration control strategy gives lower effluent concentrations
and requires lesser total aeration energy compared to a traditional constant DO control. It was
also highlighted that tuning of the aeration volume controller must be done with care.
____________________________________________________________________________________________

A risk assessment model for settling problems of microbiological origin in activated sludge
systems was proposed by Comas et al. (2008). Common operational problems are caused by
microorganism population imbalances between filamentous and floc-forming bacteria leading
to filamentous bulking, filamentous foaming and deflocculation. A fuzzy rule-based system for
risk assessment for settling problems was tested using BSM1. A knowledge related to risk of
filamentous bulking proliferation was synthesized into a decision tree; moreover, each of the
three branches of the decision tree evaluates one of the three main causes: low DO
concentration, nutrient deficiency and low F/M (food-to-microorganism) ratio or substrate
limiting conditions. Seven variables can be used by the decision tree as indicators to assess risk
of filamentous bulking: SRT, DO, F/M (removed), F/M (fed), BOD/N, BOD/P and Ss (readily
biodegradable substrate concentration). The risk estimation implies three main steps:
fuzzification, fuzzy inference of the risk through Mamdani approach, and defuzzification of the
output variable. Two case studies in modelled activated sludge environment were performed
and thus detection of microbiology-related separation problems illustrated. The output from the
proposed risk assessment system can be used as plant performance criteria together with the
usual criteria: operational costs and effluent quality.
BSM2 applications
Lately, IWA/COST Benchmark Simulation Model No. 2 (BSM2) protocol has been used in
plant-wide studies on control strategies and performance evaluation. The plant used in the
protocol is in Figure 2.
Effects of different control strategies can be evaluated using multi-criteria decision analysis
(Flore-Alsina et al., 2008). Evaluation of six WWTP control strategies was demonstrated using
BSM2 as a case study; in particular, environmental, legal, technical and economic objectives
were taken into account. Plant performance evaluation was done using data of one week
dynamic simulation. Multi-criteria evaluation was performed both without uncertainty and with
uncertainty by means of model input probability distributions. Three classes of uncertainty were
distinguished for stoichiometric (5% upper and lower bounds), kinetic (25%) and influent
fraction (50%) parameters. Different control strategies were noticed beneficial regarding to
different objectives considered. The results were analyzed using several descriptive statistical
tools and it was possible to see how different uncertainties affected the different outcomes.
Multi-criteria decision analysis results in a more transparent decision making process, improved
understanding on the process and the trade-offs between different objectives; additionally,
uncertainty analysis allows identification of potential problems at WWTP early on.
Multivariable statistical techniques were used in analysis of plant-wide WWTP control

strategies; furthermore, the methods used were cluster analysis (CA), principal component
analysis/factor analysis (PCA/FA) and discriminant analysis (DA). BSM2 protocol was used in
evaluation of 12 different control strategies. The data generated during the last 364 of the
simulation were used for the evaluation. Variables relating to economical, environmental,
technical and legal aspects were used for evaluation criteria. First, CA indicated that there were
four main types of control strategies. Next, PCA/FA showed the main correlations between the
evaluation criteria and the control strategies influencing those criteria. While there were 26
original variables, five principal components explained 96% of the total variability. Then, DA
showed that only six parameters were useful to discriminate within the classes obtained by CA.
To summarize, the combination of these multivariable statistical techniques improved
____________________________________________________________________________________________

significantly the accessibility of information needed for effective evaluation control strategies
(Flores et al., 2007).
Real plant control examples

A methodology to estimate the costs and benefits of advanced control for WWTPs was studied
by Devisscher et al. (2006); the methodology is called MAgIC (matrix for advanced
instrumentation and control). Wastewater treatment process was modelled by implementing
Simulink and the modelling framework was based on ASM1 model. The main cost categories
considered were: (1) aeration energy, (2) pumping energy, (3) mixing energy, (4) chemicals,
and (5) sludge treatment and disposal. Because of few quantitative data for influent
composition was available, a software synthetic influent generator was built and a daily pattern
with variable amplitude was added to the flow rate data. Operational data was taken from daily
reports of the plant consisting of daily incoming flow rates, operational settings such as DO set-
points, waste sludge flow rates, recirculation fractions and denitrification percentages. Both
manual and on-line controls were simulated; furthermore, short response time of the on-line
control compared to an ideal manual control was noticed to decrease aeration energy and
precipitation chemical flow rate while at same time to improve the effluent quality. Saving in
aeration energy by advanced control ranged from 10 to 20%, and saving in chemicals dosing up
to 30%. It was estimated that on-line control would become cost-effective at plant sizes above
50 000 p.e., if a payback period shorter than two years was required. The methodology has been
applied to four WWTPs in Belgium; the plants had been in operation for at least two years in
order to have a sufficient amount of data.
A hierarchical two-level control to track DO reference trajectory in activated sludge process

was proposed and investigated by Piotrowski et al. (2008). A hierarchical multilayer control
structure presented in Figure 15 utilizes multiple time scales in the plant dynamics; the fast
control sub-layer (FCS) and the medium control sub-layer (MCS) for aeration control were
considered in the research. An aeration system is a complicated hybrid nonlinear dynamical
system with faster dynamics than the internal dynamics of the DO at a biological reactor. The
objectives of the control are to force DO concentrations in the zones of activated sludge process
to follow the prescribed references and also to minimize the associated electricity cost due to
blowing air to the reactor. The control handles in typical aeration systems are: blower structure
on/off, blower speeds and the angular positions of the throttling valves. In the aeration control
structure the lower level controller (LLC, operates on fast time scale) uses the aeration control
handles in order to produce the airflow trajectories that follow the trajectories prescribed by
upper level controller (ULC, operates on slow time scale) and to minimize the electricity costs
due to aeration. There are minimum and maximum pressure constraints for airflow blown. Two
types of blowers can be used: fixed-speed motors and variable-speed motors. A hybrid
nonlinear model predictive control (HNMPC) was applied for LLC because of nonlinearity of
the input-output model.
The two-level DO control structure was tested at Kartuzy WWTP in northern Poland;
specifically, the treatment process used was UCT (University of Cape Town) process. The
process was modelled applying ASM2d model, which was calibrated on the plant operation
data; the model was implemented in a simulation package SIMBA and calculations were
carried out in Matlab and General Modelling Environmental Systems (GAMS) environment.
The HNMPC optimization task at the LLC was solved using GPLEX solver. It was concluded
that the hierarchical controller optimizes the operation costs and hence does not throttle the
____________________________________________________________________________________________

airflow by closing valves more than necessary. As a result, the computing time needed by the
controller to generate the control actions was noticed to be small and real-time implementation
at the WWTP to be very feasible (Piotrowski et al., 2008).
Figure 15. Information structure of the two-level hierarchical controller for DO optimised
tracking (Piotrowski et al., 2008).
A supervisory control strategy for Galindo-Bilbao WWTP (Spain) was designed and tested
combining model simulations, pilot-plant experimentation and full-scale validation (Ayesa et
al., 2006). The control strategy combines three complementary control loops in order to
optimize nitrogen removal in a pre-denitrifying activated sludge plant. The first controller
maintains the selected 24-h-averaged ammonium concentration in the effluent by choosing the
most appropriate DO set-point in the aerobic reactors. 24 hours mobile-averaged window is
used to filter the typical variation in the effluent concentration created by the daily load
variation. The second control strategy aims at the optimum use of the denitrification potential of
the plant by selecting the most appropriate nitrate recycle flow. The objective is to maintain a
low and stable concentration of nitrates at the end of the anoxic zone; additionally, the control
is based on the instantaneous value of the nitrates in the anoxic zone and does not need a
mobile averaged filter. The third control loop maintains the selected 24-h-averaged value of the
total mass of suspended solids in the biological reactor via automated manipulation of the
sludge wastage rate.
____________________________________________________________________________________________

Figure 16. Schematic description of supervisory control loops for (a) DO set-point control, (b)
internal recycle control and (c) sludge wastage control (Ayesa et al., 2006).
Modelling and simulation were performed using ASM1 model and WEST simulation platform;
moreover, the optimum operational strategies were designed using model-based optimization
algorithms. The controllers were designed and verified using model simulations and pilot-scale
experimentation. Similarly, the full-scale implementation and validation of supervisory control
strategies (Figure 16) was done at one treatment line of the Galindo-Bilbao WWTP. A
simultaneous reduction of both 2.0 mg/l of effluent nitrates (and consequently in total nitrogen)
and a fall of 15-20% in aeration has been obtained in full-scale operation, corroborating the
results estimated by simulations (Ayesa et al., 2006).
3.4 Plant-wide control

Plant-wide control can be defined as the combined sewer network and WWTP control and
operation (Olsson et al., 2005). By integrating controls of sewer network and WWTP capacities
of both systems can be used more efficient. Especially at the seasons with heavy rainfalls and
storm waters WWTPs are often overloaded, which may lead to increased nutrient
concentrations in the effluent and possible washout of the activated sludge organisms. Aim of
the integration is minimized impacts on the receiving water.
By using plant-wide control the storage capacity of the sewer system (and possible equalization
basins) is utilized. This does not only help operation at times of high flow rates; when the
storage capacity is properly managed, different wastewaters are better mixed and the load to
WWTP is more equal, which easies operation. On the other hand, if WWTP and sewer network
are operated separately, the plant influent flow rate is an external disturbance. By combining
control of WWTP and sewer network the influent flow becomes a manipulated variable and the
external disturbances are located further upstream. Integrated computer system gives the
necessary information infrastructure for plant-wide control. A plant-wide control system
____________________________________________________________________________________________

assumes that unit processes are controlled locally; however, it does also consider the
interactions between different parts of the plant (Olsson and Jeppson, 2006).
Plant-wide modelling has also been developed and done research on. In the models there are
sewer network, wastewater treatment plant and receiving water or two of the aforementioned.
Calibrating plant-wide models is challenging; the individual submodels have first to be
calibrated and verified and after which the complete model has to be tested and verified for
different situations. In Olsson and Jeppsson (2006), River Water quality Model no. 1
(RWQM1) has been developed with the intended aim of compatibility with the ASM models.
Schtze et al., (2004) stated that the systematic development, evaluation and tuning of
integrated control procedures is expected to be done with integrated simulators are currently in
full development. In their work a global optimal predictive real time control (RTC) system was
implemented in Quebec Urban Community. The system consists of finding the flow set-points
that minimize the value of a multi-objective function, with respect to physical and operational
constraints. The control objectives in the RTC system are in decreasing order of priority: the
minimization of overflows, the maximization of the use of treatment plant capacity, the
minimization of accumulated volumes and the minimization of variations of the set-points. The
sewer network distributed control procedure is divided in three hierarchical levels: level 1
consists of local control of the actuator; level 2 includes several level 1 stations; and the global
optimal predictive RTC represents the level 3. In particular, RTC control system uses flow
monitoring, water level data, rainfall intensity data, radar rainfall images and 2-h rain
predictions. The optimization problem is solved every 5 min control time step by a nonlinear
programming algorithm. By optimizing the use of two tunnels and capacity of the Westerly
WWTP, RTC achieved 70% reduction in overflow volume in 2000. The costs of planning a
RTC can be quite high, but it can help to prevent building new tanks or extending capacity of
existing infrastructure at even higher costs, thus often resulting in significant savings. The RTC
system has to adapt to different conditions and varying rainfall intensities and space and time
distribution. Also, the data has to be synchronized and updated more quickly for control than
for supervisory purposes only, especially in sewer systems with short runoff concentrations and
flow times.
Control structure and algorithms for optimizing control of integrated wastewater treatment
plant-sewer systems (IWWTS) were designed and implemented at Kartuzy WWTP in northern
Poland by Brdys et al. (2008). Control structures were combined with multivariate optimizing
algorithms. Model predictive control (MPC) was used and control strategies for three
operational states were designed: normal, disturbed and emergency operational state. During
high hydraulic loads the aim of the control system is maximizing utilization of the hydraulic
capacity of the sewer and equalization tanks and maximizing hydraulic capacity of the WWTP.
The treatment process used at Kartuzy WWTP is UCT (University of Cape Town) process. In
the multilevel hierarchical structure of the control system (Figure 17) there are three different
levels: a supervisory control layer (global knowledge of current activity in the entire system),
an optimizing control layer (control objectives split in three sub-layers; slow, medium and fast
process dynamics time scales) and a follow-up control layer (in which the simple controllers
following the set-points are located). Firstly, Slow Control Sublayer handles objectives over a
horizon of a week up to several months; manipulated variables are sludge retention time, mass
sludge and pumping in/from equalization and septic tanks. Secondly, Medium Control Sublayer
handles objectives over a horizon of a day; manipulated variables are DO concentrations,
____________________________________________________________________________________________

recirculation flow rates and chemical precipitation. Finally, Fast Control Sublayer operates
objectives over a horizon of one hour; it is responsible for effluent quality during heavy rains
and of short duration events, actuator constraints and meeting demands on desired DO
concentrations prescribed in the previous sublayer. A grey box model, which results from a vast
simplification of ASM2d model, was used in the research for the predictive control. The
different control strategies for various operational stages were switched softly. In fact, the main
objective of control strategy CS1 is to minimize an overall operation costs. Moreover, for CS2
the main objective is to minimize the discharged pollution load and for CS3 to avoid or
minimize overflows. In the simulation of the MPC SIMBA 4.0 package was used as a real plant
simulator while sequential quadratic programming (SQP) solver from TOMLAD Optimisation
was applied to carry out the optimization tasks. For the simulations a 6-day long testing input
was designed and controller was successfully validated.
Figure 17. Hierarchical intelligent control structure (Brdys et al., 2008).
____________________________________________________________________________________________

Chapter 4. Sensors and data analysis
Until very recently instrumentation was one of the main bottlenecks in improving wastewater
treatment automation and consequently control and monitoring were complicated tasks in the
plant. Mainly due to the fact that during the last years WWTPs have been upgraded from being
relatively simple mechanical/biological plants for removal of organic matter and suspended
solids to being complex processes with physical, chemical and biological units, most of the
plants implemented modern SCADA (supervisory control and data acquisition) systems for
substituting old techniques for data acquisition and control. Even if the control of the plant is
still based on simple schemes, the instrumentation is no longer being considered a bottleneck
for monitoring WWTPs (Olsson et al., 2006).
Information collected in SCADA system of WWTP can be reused in soft sensors for e.g.
instrument monitoring, fault detection and redundancy control. The aim of this Chapter is to
give a brief but yet exhaustive overview on the used instrumentation and data analysis done in
WWTPs.
4.1 Sensors and Analysers
On-line sensors
As stated by Olsson and Newell (1999), the sensors are the eyes and ears of the control
systemand it is quite obvious that process control needs sensors and analyser for continuous
on-line implementation. Common sensors are reported in Table 1, their use is increasing for
advanced control of WWTP, and this gives rise to an important improvement in operating
safety and better operational economy (Olsson et al., 2004).
Table 1. Commonly uses measurements in WWTPs (Olsson et al., 2004).
Flow Sludge concentration
Level and pressure Sludge blanket
Temperature Nutrients
pH Total N and P
Redox Organic matter with UV absorbance
Conductivity Fluorescence
Oxygen Biogas
Turbidity
In particular, nutrient sensors technology is based on automated laboratory methods, it requires

sample flow without suspended solids which represents in some extend the weakness in the on-
line measurement: sampling and possible pre-treatment of the sample stream. The development
of nutrient sensors is based on the solution of these key aspects, focusing also on the reduction
of chemical consumption, leaving still room for improvements.
Soft-sensors
When nutrients analyser or generally speaking some needed measurements are not available on-
line, in a successful manner, they can be estimated a soft-sensor, which represents a
combination of robust hard-sensors and a mathematical model defined to reconstruct the time
evolution of the unmeasured states. In this way, we can build very powerful monitoring tools,
____________________________________________________________________________________________

which can be used to follow the time evolution of variables that are not accessible on-line. Such
tools can also be used for diagnosis about the operation of the plant and help the operator or a
supervision system to take the appropriate actions to maintain the process in a good operating
conditions, diagnose possible process failure or prevent accidents.
A software sensor can be defined as an algorithm built from a dynamical model of a process to
estimate on-line unmeasured variables and/or unknown (or poorly known) parameters (e.g.,
specific reaction rates, or some other kinetic or yield coefficients) from few measurements
available on-line (typically, flow rate, nutrient concentrations, turbidity, pH, etc.). In that sense,
these tools can be viewed as sensorsbased on an algorithm (software): for this reason they
are called software sensors or soft-sensors. They essentially refer to the state estimation
problem of determining the values of the states variables.
In order to describe the basic concepts of a state estimator, we consider the simple case of a
process model (without any measurement errors or process noise) as expressed in the state
space, time invariant form:
x& = Ax + Bu
(4.1)
y = Cx
The matrices A, B, C contain the characteristic parameters of the system, u is the vector of the
inputs or forcing functions, x are the state variables and y are the output observations.
If the system in 4.1 is observable, on-line estimates of the states x are obtained from the
following observer equation 4.2 in which a driving term is included to minimizing the
observation error ( e = y y) between measured values y and model prediction y= Cx:
x&= Ax+ Bu + K(y y) (4.2)
Estimates of the states are therefore obtained by simply integrating equation 4.2 and the design
of the observer reduces to the adequate choice of the matrix K, the gain matrix. The standard
approach starts from the desire to minimize the observation error. The aim is reduced to the
problem of designing the gain matrix in such a way that the observation error decreases in a
desirable way.
Depending on the way to choose the gain matrix K, several examples of software sensors can
be found in literature. In particular, there exist four fundamental approaches to observer design
(Alvarez, 2000) for nonlinear system:
Extended Kalman filter (Kalman, 1960): the most widely used state estimation technique in
chemical engineering. Its design is simple but lacks in both stability criteria and systematic
tuning procedures.
Geometric observer (Kremer, 1987): which guarantees robust stability with linear input-
output errors dynamics.
High-gain approach (Gauthier, 1992): which guarantees stability, but has a complex tuning
procedure.
Sliding mode (Slotine, 1987): which guarantees robust stability, but has an elaborate design.
____________________________________________________________________________________________

Many soft-sensor applications in wastewater treatment can be found in literature. For example,
Lindberg (1997) proposed on-line methods for estimating the time-varying respiration rate and
the nonlinear oxygen transfer function from measurements of the dissolved oxygen
concentration. Also a work presented by Sotomayor et al. (2002) deals with oxygen estimation:
they considered the design of a soft-sensor for on-line estimation of the biological activity of
micro-organisms in an ASP, which are intimately related to the dissolved oxygen concentration.
Benazzi et al. (2005) proposed a soft-sensor implementation based on an EKF for the on-line
tracking of the total suspended solids. Aguado et al. (2006) compared different predictive
models for nutrient estimation in a sequencing batch reactor (SBR): principal component
regression (PCR), partial least squares (PLS), and artificial neural networks (ANNs), in order to
find the most feasible soft-sensor for the SBR. ijl and Lumney (2006) investigated a flow
control throughout Rya WWTP in Gteburg (Sweden). The controller was implemented to
reduce internal disturbances and to provide a firm foundation for more advanced process
control. No flow measurements was available in all the points they were needed for control; for
those points flow rate was estimated where necessary by simple flow balance models mainly
based on available on-line instrumentation. For pumping stations, soft-sensors were based on
lift height and motor frequency or power. In this case, soft-sensors were also used for alarms,
the flow control system for operating bypass water gate and adjustment of recirculation flow
set-point of trickling filters. Soft sensors have been noticed to be robust and to reliably indicate
on-line measurement deviations. Nonlinear geometric observer was used for nutrient
component estimations by Mulas (2006), in order to predict the dynamic behaviour of nitrogen
compound in activated sludge process for control purposes.
Predicting the wastewater characteristics

Variations in wastewater flow rate and concentrations of components occur daily, monthly and
yearly; in certain patterns the variation can be described by time series model. A time series
model to forecast influent flow rate, temperature, COD, NH4-N and PO4-P concentrations for
seven days in advance was developed in South Korea (Kim et al., 2006). Three different
methods were tested for developing a simple forecasting model according to the number of
models and forecasting steps (Figure 18). Field plant operation data of D-city WWTP for 150
days and 100 days were used for model development and model validation respectively. The
influent data was not measured every day; in fact, the measurement intervals were 2-4 days on
average and 10 days at a maximum. Missing values were interpolated with the spline method
for application to the time series analysis. ARIMA (autoregressive integrated moving average)
analysis was also performed to generate re-interpolated values of influent characteristics
considered and a linear regressive model was developed. In conclusion, the method No. 3 (One
model and one-step-ahead forecasting) was considered to be the best of the three methods
tested since forecasting did not result in any accumulated error and the coefficients estimation
was simple and easy.
Dellana and West (2009) have made a research on comparison of multi-period predictive ability
of linear ARIMA models to nonlinear time delay neural network (TDNN) models in wastewater
treatment applications. In the study artificially generated data sets were used to simulate
wastewater process characteristics as well as real-world wastewater data sets. TDNN was
clearly superior for single period forecasting. However, when complexity was increased the
TDNN did not have an accuracy advantage over ARIMA. TDNN was more accurate for single
period and multi-period predictions for measures of nitrogen, phosphorous and BOD.
____________________________________________________________________________________________

Figure 18. Procedure for development of time series models with three different methods (Kim
et al., 2006).
As plant-wide modelling and benchmarking simulations will apparently be important research
and development methods in the near future, there is still a major challenge for plant-wide
control in real plants: predicting the influent flow rate. For this reason, using data from
pumping stations in the catchments area of the plant instead of using rain gauges a major
uncertainty in the calculations is avoided (Olsson and Jeppson, 2006).
4.2 Data analysis and fault detection

Sometime having a lot of measurement data available means that there is the need to extract
relevant information. In this case, instrumentation has to be combined with adequate data
screening, measurement processing and more or less sophisticated extraction of feature from
the measurements, since the today problem is not the lack of data, but data that reveals adequate
information.
Data screening
The possibility to finely perform on-line monitoring and process control depends on the quality
of data; for this reason, data screening represents a very important task. Data screening methods
provide the researchers, as well as the plant operators, essential means to detect potential data
problems by identifying data entry errors, missing values, possible outliers, non-normal
distributions, etc. A compressive but yet complete review on data screening methods used also
in wastewater application can be found in Olsson and Newell (1999) and later in Olsson et al.
(2004).
Data reconciliation
Measured data of WWTPs often contain errors, which can be reduced by using data
reconciliation. This is essential before using data for modelling, process evaluation, process
design and benchmarking purposes. Application of mass balances in difficult because of
____________________________________________________________________________________________

dynamic nature and variability of the influent load; additionally, accuracy of design parameters
like sludge retention time (SRT) and internal conversion rates is highly important.
For example, mass balances of Deventer WWTP in the Netherlands using biological
phosphorous removal process were studied by implementing gross error detection and data
reconciliation software Macrobal (Puig et al., 2008). The WWTP was intensively sampled
during three separate days during dry weather flow conditions. Overall flow and total
phosphorous (TP) balances contained significant errors caused by e.g. measurement errors,
which could be fitted using reconciliation and balancing software; also standard deviations
were highly reduced. The SRT was calculated in four different ways; the best results were
obtained way by using the TP balance based on TP leaving the treatment process. When using
COD measurements, the non-measured oxygen consumption was noticed to make
reconciliation of the balance impossible.
Fault detection and diagnosis

Closely linked with data acquisition is the detection of faulty behaviour from process data.
Fault detection and diagnosis (FDD) methods are a family of approaches finalized to detection
and location of faults on processing systems. The term fault is generally defined as an
unpermitted deviation of at least one characteristic property of the system from the acceptable,
usual, standard condition (Isermann, 2006). The main aim of FDD methods consists in
determining the type, size and location of the most possible fault, as well as its time of detection
in order to improve product quality, plant efficiency, as well as ability to meet environmental
and safety constraints. Research in this field has been very active since the 1970s and numerous
fault diagnosis methods are reported in the industrial and academic literature.
Interest in monitoring systems for FDD in biological processes and, specifically, WWTPs has
started around 30 years ago with an early application of model identification and diagnosis by
Beck et al. (1978) based on a fuzzy control. Since then, several alternative approaches were
presented, mainly based on statistical considerations. Fuente et al. (1996) compare a number of
established techniques for estimating the process parameters when a fault occurs in a simplified
nonlinear model for a real plant. Yoo et al. (2003) proposed the identification of different
operating conditions and discrimination between sustained and simple faults by using data-
based methods for dimensionality reduction and clustering. In a similar fashion, Lee et al.
(2004) focused on the identification of faulty sensors and the reconstruction of the
measurements by estimating the process dynamics; here, a dynamic principal component
analysis (PCA) was successful in effectively detecting the faults, especially when the abnormal
condition was caused by a single sensor failure. In the more recent work, Ciappelloni et al.
(2006) proposed a combination of model- and data-based techniques to describe a pilot-scale
aerobic sequence batch reactor (SBR); PCA is again used to describe different process cycles
and a decision algorithm was developed to classify different process faults. The SBR was also
studied by Wimberger and Verde (2008), with the goal of evaluating the detectability and
isolability for analytical- and signal-based fault detection and isolation. A different approach
was proposed by Nejjari et al. (2008), where an extended Luenberger observer is adopted for
the residual generation of sensor faults. Zumoffen and Basualdo (2008) focused on the need for
an efficient fault tolerant control on WWTP. Nevertheless, the data-based approach to fault
detection methods can achieve only limited validity, mainly because vast amounts of data are
required but not necessarily available in a WWTP. Essentially, for these reasons the causal fault
diagnosis approach is believed to have a promising potential.
____________________________________________________________________________________________

Part II
ICA situation at the Finnish Wastewater Treatment
Plants
____________________________________________________________________________________________

State-of-the-art surveys on ICA at WWTPs have been performed over the years with the
perspective of different countries. Starting with one of the first overviews of ICA in the
Scandinavian countries by Olsson et al. (1998) and at the same time by Garrett (1998) in the
United States, interest in the implementation of automation in WWTPs has been progressively
growing. An international survey was provided by Ingildsen (2002), giving an interesting
picture of the actual utilization of sensors and controls in the plants based on key performance
indicators. Jeppsson et al. (2002) took the point of view of European country conditions, where
the focus was on the level of instrumentation used in medium-sized plants (>50 000 p.e.) for
on-line control. Lately, this situation has been summarized and updated by Olsson et al. (2005),
whose main conclusions were that a well established level of automation based on physical
variables and basic control of dissolved oxygen has been reached, while control based on more
advanced sensors is still in the initial stages.
Similarly, the aim of this Part is to review the current status of ICA in municipal WWTPs in
Finland. The method of investigation is based on a questionnaire including key elements
regarding plant design, operation and utilization of ICA, and operatorsopinion which was sent
to large (> 100 000 p.e.), medium-sized (30 000 100 000 p.e.) and small WWTPs in Finland.
Furthermore, on-site surveys of some of the most representative WWTPs were made.
____________________________________________________________________________________________

Chapter 3. ICA at Finnish ASPs
In order to investigate the situation of instrumentation, control and automation in Finland, a

questionnaire was carefully prepared in co-operation with Finnish wastewater treatment experts
and by utilizing the information of a WWTP survey previously conducted in Finland (Kangas,
2004). In the questionnaire, key aspects regarding plant design, operation, and a more specific
part on ICA were included. ICA questions concerned for instance sensors, automatic analysers
and their use in on-line control, type and usability of different controls, advanced controllers
and process modelling, as well as the plant operator's attitude towards ICA. The questions also
concerned the configuration and operation, removal requirements, industrial wastewaters,
wastewater temperature, chemical use and electricity consumption. Further, the major problems
and future expectations for WWTPs were queried. Altogether there were questions on thirty-
one topics in the questionnaire, some of which were divided in to several sub-questions. The
answers of the questionnaire concern either year 2006 or 2007.
Twenty-four of the investigated plants answered the questionnaire, making a response rate of
70%, and nine of them were visited (three of the plants are built inside of the bed rock). Nine
plants were chosen as a representative group of Finnish municipal WWTPs of different scales,
and in-situ investigations were organized.
3.2 Plant configurations and operation

The design of the plant has consequences for the plant efficiency and performances, and for this
reason, plant design questions were investigated first. All of the WWTPs considered in the
survey consist of activated sludge processes, with different configurations and basin shapes,
where the main objective is total nitrogen removal (in 14 plants), ammonia removal (in five
plants) and biological phosphorus removal (in two plants), as schematically reported in Figure
19. Five of the plants have no requirement for nitrogen or ammonia removal. The WWTPs
studied have been in operation from 7 to 54 years; however, all the plants excluding the latest
one have been renovated during this decade.
16
14
Quantity of WWTPs
12
10
0
Only org. NH4-N removal Total N removal Biological P
matter removal removal
Figure 19. Quantities of WWTPs designed for different substance removal.
Typically, the wastewater treatment line of a Finnish WWTP consists of screens, sand trap,
primary clarifiers, activated sludge basins and secondary clarifiers. Moreover, some of the
plants have tertiary treatment, and equalization basin or middle clarifier. Flotation is the most
common tertiary treatment unit in use at four of the WWTPs included in the study, while post-
____________________________________________________________________________________________

filters are used as a tertiary treatment process at two plants. The biological treatment process
configuration varies in the different plants as shown in Table 2. Pre-denitrification and
simultaneous nitrification and denitrification being the most common process types, the number
of biological treatment lines varies between 1 and 10 with an average 3.4 lines; however, the
number of primary and secondary clarifier lines is not necessarily the same as the number of
activated sludge lines.
The average design flow rate at WWTPs is 38 300 m3/d and the average maximum design flow
rate is 2 980 m3/h. The average current flow rate at the WWTPs considered is 29 200 m3/d;
however, the flow rates of the plants differ substantially with the range of average flow rates
from 2 150 260 000 m3/d. In addition, the proportion of current flow rate to design flow rate
varies from 35 to 104%, the average being 69%.
Table 2. Number of different nitrogen removal processes.

Type of treatment process Number
Pre-denitrification 6
Simultaneous nitrification / denitrification 6
Post-denitrification 2
Alternate nitrification / denitrification 3
Only nitrification 4
The considered average sludge production at the WWTPs is 1.2 kg TS/kg BOD7, the standard
deviation being 0.53 kg TS/kg BOD7. The water content of sludge varies from 68 to 94% while
the average value is 76%. The sludge productions (tn/a) as dry solids and flow rates at WWTPs
are presented in Figure 20.
20000 300000
18000
Sludge production
Sludge production, tn/a
250000
16000
Flow rate
Flow rate, m3/d
14000
200000
12000
10000 150000
8000
100000
6000
4000
50000
2000
0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Figure 20. Sludge productions and flow rates at WWTPs.

The mean sludge age used in plant operation during wintertime is 13.5 d and during
summertime 9.5 d. Additionally, 10 of the plants are operated according to target sludge age
and 11 according to target sludge concentration. Regarding the share of industrial wastewater
(from e.g. food, paper, chemical and textile industries) 10.5% of influent flow rate and 19.8%
of influent load were found to be average values in the considered plants. The average COD/N
ratio of the influent is 11.9, which is considered low for a denitrification process without the
use of an external carbon source.
____________________________________________________________________________________________

The usual problematic conditions at Finnish WWTPs are snowmelt and heavy rain; in such a
situation, the influent flow rate can be too high, and for that reason also bypasses controlled
manually by the plants operators are common. In fact, at nine of the investigated plants the
biological part of the treatment process was bypassed during last year, whereas the whole
process was bypassed at ten plants, and only at six of the WWTPs were no bypasses done. The
removal requirements of 20 WWTPs were always fulfilled during the year of the questionnaire
survey, while the regulations of three plants were violated. When the whole process is not
bypassed wastewater is treated e.g. in part of the pre-settling lines with precipitation chemical
or in tertiary treatment unit such as flotation.
Energy and chemical consumptions

The operation of the plant is associated with various costs, such as chemical and energy
consumption. The consumption of precipitation and alkalinity chemicals as well as an external
carbon source varies from plant to plant. The average dosages of the most used chemicals at the
investigated plants are presented in Figure 21a. The most commonly used precipitation
chemical in Finnish WWTPs is ferrous sulphate, which is in use at 14 of the plants with an
average dosage of 128 g/m3, whilst ferric sulphate is used in nine plants. In addition to the
precipitation chemicals shown in Figure 21a, also polymer is used for precipitation at four
plants and aluminium chloride at two plants. Polymers are fed into secondary clarifiers and
used together with ferrous or ferric sulphate.
Electricity consumption, GWh/a
16 250 45 300000
Quantity of WWTPs
14 Quantity 40
Dosage, g/m3
200 Electricity consumption

12 250000
Dosage 35 Flow rate
Flow rate, m3/d

10 150
8 30 200000
6 100 25
150000
4 20
50
2 15 100000
0 0
10
Methanol
sulphate
sulphate
50000
Ferrous
hydroxide
carbonate
Calcium
Sodium
Ferric
5
0 0
1 3 5 7 9 11 13 15 17 19 21 23
a) b)
Figure 21. Quantity of WWTPs using different chemicals and average dosages of the
chemicals (a). Electricity consumptions and flow rates at WWTPs (b).
From Figure 21a, it can be seen that calcium hydroxide is the most used alkalinity chemical (12
plants); sodium carbonate is used at three of the plants; in addition, methanol is used as an
external carbon source at four of the WWTPs. At most of the plants with a total nitrogen
removal process an external carbon source is not used; instead, several of the plants are able to
utilize carbon-rich industrial wastewaters from, e.g. breweries or dairies as a carbon source for
denitrification. The range of methanol dosage is 23 56 g/m3 the average dosage being 35 g/m3.
At the WWTP with the lowest methanol consumption there is a post-denitrification process.
The electricity consumption per influent flow rate ranges from 0.17 to 1.00 kWh/m3.
Additionally, seven of the WWTPs were able to specify the amount of electricity consumed in
aeration, with the average share being 43.1% of total electricity consumption; furthermore, the
average consumption of the biological part of the plant is 54.6% of total electricity consumption
at five plants able to define the number. The average consumption of sludge treatment of the
total electricity consumption at eight of the plants is 5.8%. Six of the plants also produce
electricity on-site using biogas derived from sludge digestion; specifically, on average they
____________________________________________________________________________________________

produce 34.8% of the electricity consumed at the WWTP. The highest electricity production
rate among the plants considered is 49% of the electricity consumed. Electricity consumptions
and flow rates at WWTPs considered are presented in Figure 21b. It can be seen that there are a
few plants with substantially different electricity consumption. The reasons for this are various;
e.g. plant No. 3 in Figure 21b has only an ammonia removal requirement, plant No. 9 has no
nitrogen removal requirement and plant No. 14 is a carrier process, the configuration and
operation of which differs from normal activated sludge plants.
Wastewater quality
Average concentrations of influent wastewater at sixteen the WWTPs considered are presented
in Table 3; also average values of biological oxygen demand (BOD7) and chemical oxygen
demand (CODCr) are included. The average COD/N ratio of the influent is 11.9, which is low
for denitrification process without use of external carbon source. The range of influent COD/N
ratios at varies from 8.9 to 18.4.
Table 3. Average concentrations of influent wastewater.

Substance BOD7 CODCr Tot. N NH4-N Tot. P SS
Concentration, mg/l 248 582 48.0 37.9 8.3 297
Low temperature of municipal wastewater is typical in Finland. At the WWTPs considered the
mean wastewater temperature and the average minimum temperature are 12.3C and 6.8C
respectively. In addition, the average time, when temperature of wastewater is above 12C, is
6.2 months during a year. Moreover, the average share of industrial wastewater (from e.g. food,
paper, chemical and textile industries) at WWTPs is 10.5% of influent flow rate and 19.8% of
influent load.
Employees
At fourteen of the WWTPs employees work on five days a week during the day in a normal
situation, at four plants on seven days a week during the day and at three plants on day and
night seven days a week (Figure 22). At one of the plants there is manning on 92.5 h/wk. The
number of full-time employees at the WWTPs varies from 2 to 50 with an average 9.8 per
plant. Additionally, the average flow rate per an employee is 2 450 m3/d.
Other
Day and night

seven days a week
Seven days a week

during the day
Five days a week

during the day
0 2 4 6 8 10 12 14 16
Quantity of WWTPs
Figure 22. Manning at the WWTPs.
____________________________________________________________________________________________

Wastewater quality
Average concentrations of influent wastewater at sixteen the WWTPs considered are presented
in Table 3; also average values of biological oxygen demand (BOD7) and chemical oxygen
demand (CODCr) are included. The average COD/N ratio of the influent is 11.9, which is low
for denitrification process without use of external carbon source. The range of influent COD/N
ratios at varies from 8.9 to 18.4.
Table 4. Average concentrations of influent wastewater.

Substance BOD7 CODCr Tot. N NH4-N Tot. P SS
Concentration, mg/l 248 582 48.0 37.9 8.3 297
Low temperature of municipal wastewater is typical in Finland. At the WWTPs considered the
mean wastewater temperature and the average minimum temperature are 12.3C and 6.8C
respectively. In addition, the average time, when temperature of wastewater is above 12C, is
6.2 months during a year. Moreover, the average share of industrial wastewater (from e.g. food,
paper, chemical and textile industries) at WWTPs is 10.5% of influent flow rate and 19.8% of
influent load.
3.3 Status of ICA in Finland

In this section the results of the survey regarding the present condition of modelling, monitoring
and control at the Finnish WWTPs are reported and analyzed.
Sensors
A sensor inventory was given in the distributed questionnaire and it was asked to identify the
variables continuously measured and monitored in the plant. Altogether 18 different wastewater
characteristics are measured on-line at the 24 WWTPs considered. The number of WWTPs at
which sensors and on-line analysers are used and the number at which those are used for
control are presented in Figure 23a. The percentage of WWTPs at which sensors and analysers
are considered to function properly is shown in Figure 23b. Also, a common practice is to have
a maintenance contract with the supplier of the sensors and analysers.
Dissolved oxygen (DO), suspended solids (SS), temperature, pH and level sensors are
established technology at WWTPs; the operators consider them to function well apart from the
SS and pH sensors. Presumably the reason for this is the use of SS and pH sensors in activated
sludge basins in which there is a high concentration of solid matter. SS measurements are used,
e.g. for recycle sludge pumping control. Optical DO sensors are becoming more common at
Finnish WWTPs and the plant operators find them more reliable and easier to maintain than
galvanic and polarographic DO sensors.
Fourteen out of 24 plants use automatic NH4-N, NO3-N and PO4-P on-line analysers, but the
use of nutrient sensors in control is not common even though the operators generally consider
the sensors to function properly. The nutrient sensors are mainly in use at the plants that have a
total nitrogen removal requirement. Moreover, the most modern on-line nutrient analysers at
the WWTPs visited are calibrated automatically. The usual locations of measurement for
nutrient analysers are activated sludge basin and effluent, but NH4-N analysers are also used in
other parts of the process, e.g. primary clarifiers and influent flow at some of the plants.
____________________________________________________________________________________________

Air flow rate and air pressure sensors are regular technology at WWTPs; measurements of both
sensor types are used in aeration control. Conductivity sensors are in use at 10 plants; they are
used, e.g. in monitoring industrial wastewaters and at one of the plants in predicting the
nitrogen load coming to the activated sludge basins. Turbidity, sludge blanket level and Redox
potential sensors are used in a small number of the WWTPs considered. Even so, none of the
operators of the three WWTPs at which sludge blanket level sensors are used consider them to
function properly.
25 Do the sensors in question function properly at your W WTP?

Used at WWTPs
20 100 %
Used for control 90 %
Quantity
15 80 %
70 %
n = 24 60 %
10 No
50 %
Yes
5 40 %
30 %
20 %
0 Sludge blanket level 10 %
Dissolved oxygen (DO)
Conductivity
NH4-N
Redox potential
Suspended solids (SS)
NO3-N
Influent flow rate
Temperature
Turbidity
Level
pH
Air flow rate

Air pressure
COD
PO4-P
BOD
0%
Sludge blanket level 3

Redox potential 4
Conductivity 10
Influent flow rate 22
Turbidity 5
Temperature 24
Level 17
Air pressure 21
Air flow rate 15

pH 22
DO 24
SS 22
PO4-P 14
COD 1
BOD 1
NH4-N 14
NO3-N 14
a) b)
Figure 23. Number of WWTPs using sensors and on-line analysers and their use for control (a).
Functionality and number of sensors and on-line analysers (b).
Controls
The most applied method of aeration control is DO profile control, which is used at 18 of the
plants (Figure 24). In DO profile control, the aeration basin is divided in to several zones in
which the DO set-points differ and several sensors are used for the DO concentration on-line
measurements. At five of the WWTPs, aeration control is based on one on-line DO
measurement, whereas at one plant also automatic NH4-N analyser measurements are used for
aeration control. In addition, none of the plants considered used constant air flow or calendar
control in aeration. At two of the large plants the quantity of aerated and non-aerated zones is
controlled by automation. The average DO set-point at the nine WWTPs that were visited was
2.6 mg/l.
NH4-N sensor used

for control
DO profile control
Control based on one

DO sensor
0 2 4 6 8 10 12 14 16 18 20
Quantity of WWTPs
Figure 24. Aeration control methods used at WWTPs.
The plant operators were asked about the control types (on/off or continuous controls), the
range of the controls and the functioning of the controls. The answers are reported in
____________________________________________________________________________________________

Figure 25a, while Figure 25b summarizes the functioning conditions of the controllers used.
Apart from influent and excess sludge pumping, the majority of the controls are continuous. A
pair of other controls (polymer feed, methanol feed, neutralizing influent wastewater) is not
mentioned. The plant operators consider the control ranges for most of the controls to be
suitable even if according to the their opinion, the most common problems with the control
range are with precipitation chemical feed; at one plant the range is too wide and at two plants
too narrow; also the range of aeration control valves is too wide at two plants. At two plants the
return sludge pumping control and control of the valves used in aeration are mentioned as not
functioning properly; nevertheless, generally the controls are considered to function
satisfactorily. At the majority of the plants a part or all of the precipitation chemicals are fed
flow proportionally.
Alkalinity chemical feed Does the control function properly?
Precipitation chemical feed Alkalinity chemical feed

Aeration, compressors Precipitation chemical feed
On/off
Aeration, control valve Aeration, compressors Yes
Continuous
Excess sludge pumping Aeration, control valve No
Return sludge pumping Excess sludge pumping
Influent pumping Return sludge pumping

Influent pumping
0 5 10 15 20 25
Quantity of WWTPs 0% 20 % 40 % 60 % 80 % 100 %
a) b)
Figure 25. Number of on/off and continuous controls (a). Functioning of the controls at
WWTPs (b).
The major part of the controls at the WWTPs is implemented by using basic feedback
controllers, being the tuning done from the control room by the operators. Advanced
controllers, such as adaptive controller, fuzzy logic controller and model-based controller, are
in use at six plants for different purposes such as air flow control in aeration, mass flow rate
control of return sludge, centrifugal sludge dewatering, methanol feed, and precipitation
chemical feed. Fuzzy logic is also used to predict the nitrogen load coming in to the activated
sludge basins at one WWTP.
Alarm
The alarm management was investigated and as result it was found that different levels of
alarms are taken into account, for instance indicating faults in the process equipment. Usually at
modern Finnish plants the treatment process can be monitored and controlled remotely, e.g. on
weekends, especially for alarm purposes.
Modelling
Process modelling and simulation have been used at five of the plants; three of these have their
own modelling software in use. Three of the operators answered that modelling is also used as
help for process control; at one plant there is an expert system and at the other two modelling is
used off-line in creating control strategies. The operators mentioned studying different process
operating possibilities, process design and supporting the start-up of the process as uses for
modelling software, whereas using modelling for dynamic set-point setting is considered one
possible application in plant operation in the future. The plant operators found accurate model
calibration rather challenging, which limits the use of models. In addition, the possibilities of
using model-based controllers have not yet been taken into account.
____________________________________________________________________________________________

Two of the plants at which modelling has not been used so far have considered using
modelling, but not purchasing their own modelling software (Figure 26). Additionally, another
two WWTPs have thought of using modelling and purchasing their own software. The
remaining 14 plants at which modelling has not been used have not so far considered using
modelling at their WWTPs. Two of the WWTPs considered have recently taken part in a
research project during which the plant was modelled; however, the results of the project were
not utilized at the time of answering the questionnaire.
Have you considered of using modelling or

purchasing an own simulation software?
Yes, modelling
and purchasing
a software
Yes, modelling
No
0 2 4 6 8 10 12 14 16
Quantity of WWTPs
Figure 26. Opinions on using modelling and purchasing own simulation software at the
WWTPs not been modelled so far.
Attitude towards ICA

As final questions in the questionnaire, opinions on the current status of instrumentation and
control at the investigated plant were asked to the plant operators. It can be notice in Figure 27
that the most part of them (in 13 of the plants) considers that it would probably be possible to
gain more from the current ICA equipment in use, giving a good attitude towards ICA
implementation and improvement in their own plants. Furthermore, most of the operators
moderately agreed that the importance of ICA is going to increase in the near future, but over
one third of the interviewed operators strongly agreed as reported in Table 4.
How do you evaluate the use of the current

instrumentation and automation at your plant?
It would not be possible to benefit more

from them
It would probably be possible to benefit

more from them
It would be possible to benefit more

from them
It would be possible to benefit much

more from them
0 2 4 6 8 10 12 14
Figure 27. Opinions on the current instrumentation and automation by plant operators.
____________________________________________________________________________________________

Table 5. Opinions on importance of ICA at WWTPs in the near future: Instrumentation,
control and automation will be more important at WWTPs in the near future than they are
today
Strongly agree 9
Mildly agree 15
Neither agree nor disagree 0
Mildly disagree 0
Strongly disagree 0
The interviewed plant operators give to ICA an important role in making a more efficient
management in their plants, in particular they considered that the following points can be
helpfully solved by means of ICA:
1. Prediction of wastewater flow rate and load real-time;

2. Utilization of automatic on-line analysers in control;
3. Better aeration control;
4. More reliable the on-line measurements.
In fact, infiltration into the sewage network, heavy rainfalls and storm waters are named as the
most important bottleneck in operation of the plant in four of the answers. Additionally,
maintenance of automation equipment and reliability of measurements are mentioned often.
Even though at some of small and medium-sized plants visited the knowledge of the employees
on ICA is average, there are also exceptions; at some WWTPs development of ICA and plant
operation practices is followed closely and there is interest in implementing new technology
and practices.
At two of the WWTPs the quantity of aerated and non-aerated zones is changed by automation.
Otherwise e.g. automatic control of equalization basin operation during high flow rates and
automation relating to influent pumping sequence during and after a power cut are mentioned
as unusual automation. At one of the WWTPs considered there is a prediction system for future
influent flow rate taking also weather forecasts into account.
Considerations from the WWTP visits

Nine WWTPs were visited during the project: three large, five medium-sized and one small
plant. At one of the WWTPs there was only ammonia nitrogen removal requirement and at
eight plants total nitrogen removal requirement at the time of the visit. Process configurations
were various; the most typical process among the visited plants was DN-process and other
process types consisted of e.g. post-denitrification process, AB process and ring channel
process.
At majority of them new automation has been installed during the last few years and monitoring
software had been renewed or updated. The latest automation renewals consisted of e.g. a
reporting and data recorder software. The control rooms and SCADA systems at the visited
plants were generally modern and suitable for plant operation, but also variation in the level of
____________________________________________________________________________________________

automation and also utilization of it was noted to be remarkable; additionally, as expected, at
the large WWTPs with bigger number of employees the knowledge on ICA was better and the
automation systems were updated more regularly. Several types of sensors had been renewed at
most of the plants during the last few years; furthermore, at several plants operators said that
the plants are forced to do this, because there are no spare parts available anymore for the older
sensor types.
Several causes for operational problems or limitations were mentioned, e.g. faulty design of
channels or wrong location of chemical feed. Naturally also configuration of unit processes has
a major influence on the flexibility of operation of the treatment process; moreover, this is of a
special importance in cases when a nitrification process is modified to nitrification-
denitrification process.
In the discussions had during the visits ideas for improvements for plant operation were
mentioned. According to one idea, chemicals would be useful to feed mass flow proportionally
instead of volume flow proportional feeding generally used. This could be done, if there were
more reliable on-line nutrient measurements or soft-sensor nutrient estimates of influent
wastewater quality.
At most of the plants there are cross flow ultra filters removing suspended solids before
pumping the samples to on-line nutrient analysers. The ultra filters have to be cleaned once or
twice a week and were often mentioned to be troublesome to maintain. Another more advanced
filtering solution used at one WWTP visited is an in-situ membrane filtration unit, which is
easy to maintain and routine weekly cleaning is not needed. When the in-situ filter is about to
be clogged up and needs cleaning, an automatic alarm signal is given for plant operators.
The general attitude at WWTP visits was that it is better to have less measurements, which a
reliable, than lot of measurements some of which are not reliable. The basis of use of on-line
measurements in control and monitoring is the accuracy of the measurements.
At few plants there are also new ion selective ammonium and nitrate sensors for direct
measurement. The ion selective sensors are placed into the activated sludge basin and no
filtration or reagents are needed for the on-line measurement in contrast to generally used
nutrient analysers; moreover, the location of the sensors can be easily changed according to
what are the needs for measurements.
At smaller plants visited bypasses of the treatment process is a typical problem during heavy
rains and storm waters. At larger plants the whole process is usually not bypassed but large
water flows can be treated e.g. in part of the primary clarifier lines with precipitation chemical
or in a tertiary treatment unit. At one of the visited plants a separate Actiflo process for bypass
water treatment will be implemented in the near future.
____________________________________________________________________________________________

Chapter 4. Conclusions and future research needs
An extensive survey on ICA conditions at large and medium-sized Finnish municipal WWTPs
was carried out and the following conclusions were drawn.
Twenty-four activated sludge plants answered a questionnaire including the questions on plant
design, operation and ICA and nine plant visits were made. Pre-denitrification and
simultaneous nitrification and denitrification are most common process types in use. Average
design flow rate at WWTPs is 38 300 m3/d, but the range of the flow rates is broad. The mean
energy consumption at the plants is 0.5 kWh/m3. Aeration is the biggest consumer of the
energy; the average share of aeration is 43% of total energy consumption at WWTP. The share
of the electricity produced on-site using biogas is on the average 35% of the electricity
consumed at six of the plants. The average COD/N ratio of the influent is 12.1, which can be
considered low for a total nitrogen removal process, which is in use at fourteen of the plants. At
the WWTPs considered the mean wastewater temperature is as low as 12.3C. The average
share of industrial wastewater at WWTPs is 19.8% of influent load; however, at some of the
plants there is carbon-rich industrial wastewater, which can be used in denitrification instead of
methanol used as an external carbon source at four of the plants.
Altogether eighteen different wastewater characteristics are measured on-line at the WWTPs
considered; additionally, at fourteen plants there are automatic NH4-N, NO3-N and PO4-P on-
line analysers, but use of nutrient sensors in control is not common even though the operators
generally consider the sensors to function properly. Reasons for this are that nutrient sensors
being fairly new technology at several plants and that the maintenance of nutrient sensors and
sample pre-treatment equipment is often considered troublesome. The most applied method to
aeration control is DO profile control, which is used at eighteen of the plants; at one plant also
automatic NH4-N analyser measurements are used for aeration control. Apart from influent and
excess sludge pumping majority of the controls are continuous; furthermore, generally the plant
operators consider the controls to function properly. Major part of the controls at the WWTPs is
implemented by using basic feedback controllers; moreover, advanced controllers are in use at
six plants. Process modelling and simulation has been used at five of the plants; only at three of
those there is an own modelling software in use. Modelling and simulation is quite new
technology at Finnish WWTPs, but the popularity of modelling and simulation seems to be
increasing.
The operators at thirteen of the plants consider that it would probably be possible to gain more
from the current ICA equipment in use. This could be done e.g. with better use of on-line
measurements in control and optimizing the control strategies at the plants. Infiltration into the
sewage network and the high flow rate variations caused by heavy rainfalls and storm waters
are considered as the most important bottlenecks in operation of the plant. Especially at smaller
plants the problem can be severe since bypasses happen often and there is not much that can be
done at the WWTP. If the major problem is considered to concern condition and type of the
sewer network, implementing advanced ICA technology at the plant does not improve the
situation essentially. Fortunately at majority of the plants the situation is not the
aforementioned.
____________________________________________________________________________________________

Since the results inform a European survey on the status of ICA at WWTPs larger than 50 000
p.e. (Jeppsson et al., 2002) evaluated by national experts, no significant changes have taken
place in Finland regarding types of instrumentation and control. However, the reliability and
accuracy of on-line sensor measurement has improved since execution of the European state-of-
the-art survey, which makes use of on-line measurements in control more applicable. The use
of nutrient sensors in control is apparently still rare at Finnish WWTPs even though their use
for monitoring purposes is common. Also, the popularity of dynamic process modelling has
increased during recent years and interest of the plant operators in the use of modelling is still
growing.
At new and renovated Finnish plants conventional ICA technology is relied on, apart from a
few exceptions. The controllers used are PID feedback controllers and more advanced
controllers are not often implemented. Even though the full potential of sensors and other ICA
technology is not taken advantage of at most of the plants, the general attitude of plant
operators towards ICA is one of interest and its importance in the future is understood.
Otherwise, there are considerable differences between the level of automation technology and
the knowledge of ICA at the plants.
During the next decade, new large and medium-sized WWTPs will be built in Finland. The
possibilities of ICA should be given special attention in the design of the plants in order to
optimize the operation costs. In addition, when renovating the existing plants, automation and
control should be taken into account since, e.g. manufacturers of sensors and analysers are
doing continuous development work. Advanced control strategies for nitrogen removal would
be beneficial to implement as well as to investigate possibilities of soft sensors and dynamic
modelling in the plant operation. The level of understanding of ICA varies among the operators
of the Finnish WWTPs; usually at the larger plants the job description of the personnel is more
specialized and there is also more ICA knowledge in the operation team. In addition, because of
constant development of ICA equipment and practices, the training and education of the plant
operators is important in order make better use of the ICA technology implemented at plants
and thus improve the plant operation.
In particular, the following needs have recognised as important aspect in the improvement of
the management of WWTPs:
Dynamic modelling and simulation of activated sludge plants has not been used at
many Finnish WWTPs so far, but in the future the popularity of modelling is likely to
increase. The process modelling and simulation is a practical and powerful tool for plant
design, process development and optimizing control of the plant. Modelling can be used
in supporting many kinds of research projects at WWTPs and universities. ASP models,
which are carefully built, calibrated and verified are suitable to be used as platforms for
simulated test runs with dynamic influent data. Moreover, use of COST/IWA
benchmark simulation models should also be considered when planning investigations
on control strategies and plant run optimisation.
Practical tools for predicting the load and wastewater characteristics (particularly
ammonium load and composition of COD fractions) coming to the activated sludge
basins would help the operation of plants; thus, changes in influent flow rate could be
controlled more efficiently, aeration and dosing external carbon source would be more
____________________________________________________________________________________________

easily optimized. More on-line information of the influent wastewater would be
essential in plant operation; more reliable on-line measurements and soft-sensors
combined would make also feed-forward and mass flow based control strategies more
applicable. When considering optimisation of nutrient removal, better use of on-line
measurements for control purposes is important; furthermore, either control strategies
successfully applied for real scale plants in other countries could be used also in Finland
or new strategies taking local conditions into account developed.
Developing a general architecture for fault modelling, diagnosis and isolation in

biological wastewater treatment processes would help in more fault tolerant and stable
operation of WWTPs. Typical faults in wastewater treatment plants should first be
analysed; in addition, fault detection and diagnosis methods can be applied to the
process depending on the encountered fault types.
Operator support and training services need to be paid attention to help operators
take full advantage of on-line measurements, to run the process with a cost-efficient
strategy and to manage abnormal situations. Services for the training of new operators
and updating the skills of existing operators are also required. Training services with
simulation activities will also be useful to provide for plant operators as modelling is
becoming more common in wastewater treatment and the simulators being constantly
developed.
As a typical problem at Finnish WWTPs is high flow rate peaks during heavy rains and
storm waters, the condition of sewage network should be paid attention to in order to
prevent unnecessary infiltration into the sewage network. Also, more integrated control
of sewage network and wastewater treatment process would help in control of
problematic flow rate peaks and hence to be considered as a possible future research
topic.
Apart from progress of the recent years, development work regarding to on-line
measurement equipment is still needed because of challenging conditions for reliable
measurements at WWTPs. This does not only consider sensors and automatic analysers
but also sample pre-treatment and filtering equipment, which need to be maintained.
____________________________________________________________________________________________

References
Aguado D., Ferrer A., Seco A., Ferrer J., 2006. Comparison of different predictive models for
nutrient estimation in a sequencing batch reactor for wastewater treatment. Chemometrics and
Intelligent Laboratory Systems, 84(1-2), 75-81.
Alex J., Benedetti L., Copp J., Gernaey K.V., Jeppsson U., Nopens I., Pons M.-N., Rieger L.,
Rosen C., Steyer J.P., Vanrolleghem P., Winkler S., 2008. Benchmark Simulation Model No. 1
(BSM1). Dept. of Industrial Electrical Engineering and Automation, Lund University.
Alvarez J., 2000. Nonlinear state estimation with robust convergence. Journal of Process
Control, 10(1), 5971.
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 Science and Technology, 53(4-5),
193-201.
Baruch I.S., Georgieva P., Barrera-Cortes J., de Azevedo S.F., 2005. Adaptive neural network
control of biological wastewater treatment. International Journal of Computational Intelligence
Systems, 20(2), 173-193.
Beck M.B., Latten A. and Tong R.M. (1978). Modelling and operational control of the
activated sludge process in wastewater treatment. International Institute for Applied Systems
Analysis (IIASA), Professional Paper PP-78-10.
Benazzi F., Gernaey K.V., Jeppsson U., Katebi R., 2005. On-line concentration and detection of
abnormal substrate concentration in WWTPs using a software sensor: A benchmark study. In
Proceedings of the 2nd Conference on Instrumentation, Control and Automation (ICA2005),
Busan, Korea.
Blevins T.L., McMillan G.K., Wojsznis W.K., Brown M.W. 2003. Advanced Control
Unleashed - Plant Performance Management for Optimum Benefit. ISA.
Brdys M.A., Grochowski M., Gminski T., Konarczak K., Drewa M. 2008. Hierarchical
predictive control of integrated wastewater treatment systems. Control Engineering Practice,
16(6), 751-767.
Camacho E.F., Bordons C. 2004. Model Predictive Control. Springer.
Cartensen J., Harremoes P., Madsen H. 1995. Statistical identification of Monod-kinetic

parameters from on-line measurements. Water Science and Technology, 31(2), 125133.
Chachuat B., Roche N., Latifi M.A. 2003. Reduction of the ASM1 model for optimal control of
small-size activated sludge treatment plants, Journal of water science, 16, 837-858 (In French)
Cho J.H., Sung S.W., Lee I.B., 2002. Cascade control strategy for external carbon dosage in
predenitrifying process. Water science and technology, 45(4-5), 53-60.
____________________________________________________________________________________________

Ciappelloni F., Mazouni D., Harmand J. and Lardon L. 2006. On-line supervision and control
of an aerobic SBR process. Water Science and Technology, 53(1), 169-177
Comas J., Rodrguez-Roda I., Gernaey K.V., Rosen C., Jeppson U., Poch M. 2008. Risk
assessment modelling of microbiology-related solids separation problems in activated sludge
Systems. Environmental Modelling & Software, 23(10-11), 1250-1261.
Copp J.B. (Edit.). 2002. The COST Simulation Benchmark: Description and Simulator Manual.
Office for Official Publications of the European Communities, Luxembourg.
Cristea M.V., Agachi S.P. 2006. Nonlinear model predictive control of the wastewater
treatment plant. In 16th European Symposium on Computer Aided Process Engineering and 9th
International Symposium on Process Systems Engineering. Edited by Marquardt W., Sass R.,
Pantelides C., Elsevier.
David R., Vasel J.L., Wouwer A.V., 2009. Settler dynamic modeling and MATLAB simulation
of the activated sludge process, Chemical Engineering Journal, 146(2), 174-183
Dellana S.A., West D. 2009. Predictive modelling for wastewater applications: Linear and
nonlinear approaches. Environmental Modelling & Software, 24(1), 96-106.
Devisscher M., Ciacci G., F 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 Science and Technology,
52(4-5), 215-223.
Ekama G.A., Barnard J.L., Gunthert F.W., Krebs P., Mc-Conquodale J.A., Parker D.S.,
Wahalberg E.J. 1997. Secondary settling tanks: theory, modelling, design and operation.
Technical Report 6, IWA Publishing.
Ekman M., Bjrlenius B., Andersson M. 2006. Control of the aeration volume in an activated
sludge process using supervisory control strategies. Water Research, 40(8), 1668-1676.
Ellis G. 2004. Control System Design Guide - A Practical Guide. Elsevier.
Flores X., Comas J., Roda I.R., Jimnez L., Gernaey K.V. 2007. Application of multivariable
statistical techniques in plant-wide WWTP control strategies analysis. Water Science and
Technology, 56(75-83), 75-83.
Flores-Alsina X., Rodrguez-Roda I., Sin G., Gernaey K.V. 2008. Multi-criteria evaluation of
wastewater treatment plant control strategies under uncertainty. Water Research, 42(17), 4485-
4497.
Fuente M.J., Vega P., Zarrop M.B. and Poch M. 1996. Fault detection in a real wastewater plant
using parameter-estimation techniques. Control Engineering Practice, 4(8), 1089-1098.
____________________________________________________________________________________________

Garrett M.T. 1998. Instrumentation, control and automation progress in the United States in the
last 24 years. Water Science and Technology, 37(12), 21-25.
Gauthier J.P., Hammouri H., Othman S. 1992. A simple observer for nonlinear systems.
application to bioreactors. IEEE Transaction on Automatic Control, 37(6), 875-880.
Gernaey K., Rosen C., Jeppsson U. 2006. WWTP dynamic disturbance modelling - an
essential module for long-term benchmarking development. Water Science and Technology,
53(4-5), 225-234.
Gmez-Quintero C., Queinnec I., Babary J.P. 2000. A reduced nonlinear model of an activated
sludge process. Proceedings of International Symposium on Advanced Control on Chemical
Processes, Pisa, Italy.
Grau P., Copp J., Vanrolleghem P.A., Takcs I., Ayesa E. 2009. A comparative analysis of
different approaches for integrated WWTP modelling. Water Science and Technology, 59(1),
141-147.
Hakanen J., Sahlstedt K., Miettinen K. 2008. Simulation-based interactive multiobjective

optimization in wastewater treatment. Proceedings EngOpt 2008 International conference on
engineering optimization, Rio de Janeiro, Brazil.
Henze M., Gujer W., Mino T., van Loosedrecht M. 2002. Activated Sludge Models ASM1,
ASM2, ASM2d and ASM3. Scientific and Technical Reports, No. 9, IWA Publishing.
Henze M., van Loosdrecht M.C.C, Brdjanovic D., Ekama G.A. 2008. Biological Wastewater
Treatment: Principles, Modelling and Design. IWA Publishing.
Ingildsen P. 2002. Realising Full-Scale Control in Wastewater Treatment Systems Using In Situ
Nutrient Sensors. PhD Thesis, Department of Industrial Electrical Engineering and Automation,
Lund Institute of Technology, Lund, Sweden.
Ingildsen P., Rosen C., Gernaey K.V., Nielsen M.K., Guildal T., Jacobsen B.N. 2006.
Modelling and control strategy testing of biological and chemical phosphorus removal at
Avedre WWTP. Water Science and Technology, 53(4-5), 105-113.
Isaacs S., 1996. Short horizon control strategies for automating activated sludge process. Water
Science and Technology, 34(1-2), 203212.
Isermann R. 2006, Fault-diagnosis systems: an introduction from fault detection to fault

tolerance. Heidelberg: Springer.
Jeppsson U. 1996. Modelling aspects of wastewater treatment processes, PhD Thesis,

Department of Industrial Electrical Engineering and Automation, Lund Institute of Technology,
Lund, Sweden.
Jeppsson U., Diehl S. 1996. An evaluation of a dynamic model of the secondary clarifier, Water
Science and Technology, 34 (5-6), 1926.
____________________________________________________________________________________________

Jeppsson U. Alex J., Pons M.N., Spanjers H. and Vanrolleghem P.A., 2002. Status and future
trends of ICA in wastewater treatment a European perspective. Water Science and
Technology, 45(4-5), 485-494.
Jeppsson, U. and Pons, M.-N., 2004. Editorial: The COST benchmark simulation model
current state and future perspective. Control Engineering Practice, 12(3), 299304.
Jeppsson, U., Rosen, C., Alex, J., Copp, J., Gernaey, K.V., Pons, M.-N. and Vanrolleghem,
P.A., 2004. Towards a benchmark simulation model for plant-wide control strategy
performance evaluation of WWTPs. Proceedings 6th Int. Symposium on Systems Analysis and
Integrated Assessment in Water Management, Beijing, China.
Jeppsson U., Pons M.-N., Nopens I., Alex J., Copp J.B., Gernaey K.V., Rosen C., Steyer J.-P.,
Vanrolleghem P.A., 2007. Benchmark simulation model No. 2: general protocol and
exploratory case studies. Water Science and Technology, 56(8), 67-78.
Julien S., Babary J. P., Lessard P., 1998. Theoretical and practical identifiability of a reduced
order model in an activated sludge process doing nitrification and denitrification. Water Science
and Technology, 37(12), 309316.
Julien S., Lessard P., Babary J. P., 1999. A reduced order model for control of a single reactor
activated sludge process. Mathematical and Computer Modelling of Dynamical Systems, 5(4),
337350.
Kabouris J. C., Georgakakos A. P., 1992. Accounting for different time scale in activated
sludge process control. Water Science and Technology, 26(5-6),13811390.
Kalman R. E., 1960. A new approach to linear filtering and prediction problems, Transactions
of the ASMEJournal of Basic Engineering, 82(Series D), 35-45.
Kangas A. 2004. Operation and implementations of wastewater treatment plants. Report of the
Finnish Water and Waste Water Works Association, (in Finnish).
Kim J.R., Ko J.H., Im J.H., Lee S.H., Kim S.H., Kim C.W., Park T.J. 2006. Forecasting influent
flow rate and composition with occasional data for supervisory management system by time
series model, Water Science and Technology, 54(4-5), 185-192.
Kova Z., Bogdan S., 2006. Fuzzy controller design: theory and applications. CRC Press.
Krener A., Isidori A, 1987. Linearization by output injection and nonlinear observers. Systems
and Control Letters, 3(1), 47-52.
Kynch G.J. 1952. A theory of sedimentation, Transactions of the Faraday Society. 48, 166-176.
Lee C., Choi S. W. and Lee I.B. 2004. Sensor fault identification based on time-lagged PCA in
dynamic processes, Chemometrics and Intelligent Laboratory Systems, 70(2), 165-178.
____________________________________________________________________________________________

Lindberg C.F. 1997. Control and estimation strategies applied to the activated sludge process.
PhD thesis, System and Control Group, Uppsala University, Sweden.
Lingireddy S., Brion G.M. 2005. Artificial Neural Networks in Water Supply Engineering.
ASCE Publications.
Marsili-Libelli S., 1989. Modeling, identification and control of the activated sludge process.
In Advances in Biochemical Engineering/Biotechnology. 38, 90148. Springer-Verlang, Berlin.
Mulas, M. 2006. Modelling and control of activated sludge processes. PhD thesis. University of
Cagliari, Italy.
Mulas M., Tronci S., Baratti R. 2007. Development of a 4-Measurable States Activated Sludge
Process Model deduced from the ASM1, Proceedings of DYCOPS, Cancun, Mexico.
Nejjari F., Puig V., Giancristofaro L. and Koehler S. 2008. Extended Luenberger observed-
based fault detection for an activated sludge process. Proceedings 17th IFAC World
Congress, Seoul, Korea.
Nopens I., Batstone D.J., Copp J.B., Jeppson U., Volcke E., Alex J., Vanrolleghem P.A. 2009.
An ASM/ADM model interface for dynamic plant-wide simulation. Water Research, 43(7),
1913-1923.
Olsson G., Aspegren H. and Nielsen M.K. 1998. Operation and control of wastewater treatment
a Scandinavian perspective over 20 years. Water Science and Technology, 37(12), 1-13.
Olsson G., Newell B. 1999. Wastewater Treatment Systems. Modelling, Diagnosis and Control.
IWA Publishing.
Olsson G., Jeppsson U. 2006. Plant-wide control: dream, necessity or reality? Water Science
and Technology, 33(3), 121-129.
Olsson G., Nielsen M.K., Yuan Z., Lynggaard-Jensen A., Steyer J.-P. 2005. Instrumentation,
control and automation in wastewater systems. IWA Publishing.
Peng Y., Ma Y., Wang S., Wang X. 2005. Fuzzy control of nitrogen removal in
predenitrification process using ORP. Water Science and Technology, 52(12), 161-169.
Phillips H.M., Sahlstedt K.E., Frank K., Bratby J., Brennan W., Rogowski S., Pier D.,
Anderson W., Mulas M., Copp J.B., Shirodkar N. 2009. Wastewater treatment modelling in
practice: a collaborative discussion of the state of the art. Water Science and Technology, 59(4),
695-704.
Posio J. 2002. Model predicative control. Oulu University, Control engineering laboratory,
Report B No 9, (In Finnish)
Qin S.J., Badgwell T.A. 2003. A Survey of Industrial Model Predictive Control Technology.
Control Engineering Practice, 11(7), 733-764.
____________________________________________________________________________________________

Pietrowski R., Brdys M.A., Konarczak K., Duzinkiewitcz K., Chotkowski W. 2008.
Hierarchical dissolved oxygen control for activated sludge processes. Control Engineering
Practice, 16(1), 114-131.
Puig S., van Loosdrecht M.C.M., Colprim J., Meijer S.C.F. 2008. Data evaluation of full-scale
wastewater treatment plants by mass balance. Water Research, 42(18), 4645-4655.
Queinnec I., Dochain D. 2001, Modelling and simulation of the steady-state of secondary
settlers in wastewater treatment plants, Water Science and Technology, 43(7), 3946.
Rodrguez J., Premier G.C., Dinsdale R., Guwy A.J. 2009. An implementation Framework for
wastewater treatment models requiring a minimum programming expertise. Water Science and
Technology, 59(2), 367-380.
Samuelsson P., Carlsson B. 2001. Feedforward control of the external carbon flow rate in an
activated sludge process. Water Science and Technology, 43(1), 115-122.
Samuelsson P., Halvarsson B., Carlsson B. 2007. Cost-efficient operation of a denitrifying

activated sludge process. Water Research, 41(11), 2325-2332.
Schtze M.R., Butler D., Beck M.B. 2002. Modelling, simulation and control of urban
wastewater systems. Springer.
Schtze M., Campisano A., Colas H., Schilling W., Vanrolleghem P.A., 2004. Real time
control of urban wastewater systems where do we stand today? Journal of Hydrology,
299(3-4), 335-348.
Seborg D.E., Edgar T.F., Mellichamp D.A. 2003. Process Dynamics and Control, 2nd Edition,
Wiley.
Shen W.H., Chen X.Q., Corriou J.P. 2008. Application of model predictive control to the
BSM1 benchmark of wastewater treatment process. Computer and chemical engineering,
32(12), 2849-2856.
Slotine J.J.E., Hedrick J.K., Misawa E.A. 1987. On sliding observers for nonlinear systems.
Journal of Dynamic Systems, Measurements and Control, 109, 245-252.
Smets I.Y., Haegebaert V.J., Carrette R., Van Impe J.F. 2003. Linearization of the activated
sludge model ASM1for fast and reliable prediction, Water Research, 37(8), 1831-1851.
Sotomayor O.A.Z., Won Park S., Garcia C. 2002. Software sensor for on-line estimation of the
microbial activity in activated sludge system. ISA Transaction, 41(2), 127143.
Sprandio M., Queinnec I. 2004. On-line estimation of wastewater nitrifiable nitrogen,

nitrification and denitrification rates, using ORP and DO dynamics, Water Science and
Technology, 49(1), 31-39.
____________________________________________________________________________________________

Stare A., Vre ko D., Hvala N., Strm nik S. 2007. Comparison of control strategies for nitrogen
removal in an activated sludge process in terms of operating costs: A simulation study. Water
Research, 41(9), 2004-2014.
Tchobanoglous G., Burton F.L., Stensel H.D. 2003. Wastewater engineering: treatment and
reuse, Metcalf & Eddy, Published by McGraw-Hill Professional.
Traor A, Grieu S., Thiery F., Polit M., Colprim J. 2006. Control of sludge height in a
secondary settler using fuzzy algorithms. Computer and chemical engineering, 30(8),
1235-1242.
Verbruggen H.B., Babuka R. 1999. Fuzzy logic control: advances in applications. World
Scientific.
Verdickt L.B., Van Impe J.F. 2002. Simulation analysis of a one-dimensional sedimentation
model, Preprints of the 15th triennial IFAC World Congress, International Federation of
Automatic Control (CDROM), Barcelona, Spain.
Visioli A. 2006, Practical PID Control, Springer.
Vitasovic Z.Z. 1986. An integrated control strategy for the activated sludge process, Ph.D.
Thesis, Rice University, Houston, TX, USA.
Yong M., Yongzhen P., Jeppson U. 2006. Dynamic evaluation of integrated control strategies
for enhanced nitrogen removal in activated sludge processes. Control Engineering Practice,
14(11),1269-1278.
Yong M., Yongzhen P., Shuying W. 2005. Feedforward-feedback control of dissolved oxygen
concentration in a predenitrification system. Bioprocess and biosystems engineering Journal,
27(4), 223-228.
Wimberger D. and Verde C. 2008. Fault diagnosticability for an aerobic batch wastewater
treatment process. Control Engineering Practice, 16(11), 1344-1353.
Yoo C., Vanrolleghem P.A. and Lee I.B. 2003. Nonlinear modelling and adaptive monitoring
with fuzzy and multivariate statistical methods in biologic wastewater treatment plants. Journal
of Biotechnology, 105(1-2), 135-163.
Zhao H., Isaacs S. H., Soeberg H., Kummel M., 1995. An analysis of nitrogen removal and
control strategies in an alternating activated sludge process. Water Research, 29(2), 535544.
Zhao L., Chai T. 2005. Wastewater BOD forecasting model for optimal operation using robust
time-delay neural network. In Advances in neural networks ISNN 2005 Second international
symposium on neural networks, Edited by Wang J., Liao X., Yi Z., Springer.
Zhu G., Peng Y., Ma B., Wang Y., Yin C. 2009. Optimization of anoxic/oxic step feeding
activated sludge process with fuzzy control model for improving nitrogen removal. Chemical
Engineering Journal, 151(1-3), 195-201.
____________________________________________________________________________________________

Zumoffen D. and Basualdo M. 2008. Improvements in fault tolerant characteristics for large
chemical plants: 1. Waste water treatment plant with decentralized control. Industrial and
Engineering Chemistry Research, 47(15), 5464-5481.
ijl G., Lumney D. 2006. Integrated soft sensor for flow control. Water Science and
Technology, 53(4-5), 473-482.
____________________________________________________________________________________________

Appendix A: Questionnaire
KYSELY JTEVEDENPUHDISTAMOIDEN KONTROLLI-

JA AUTOMAATIOJRJESTELMIST
Teidn yhteystietonne
Nimi Tehtvnimike
Shkpostiosoite (Ty)puhelinnumero
Tietoja jtevedenpuhdistamosta
Laitoksen nimi
Osoite
Paikkakunta Postinumero
____________________________________________________________________________________________

Shkisesti tytetyt kyselylomakkeet voi palauttaa liitetiedostoina osoitteeseen
henri.haimi@tkk.fi.
1A. Minklaisia mittauksia, mittalaitteita tai automaattisia analysaattoreita laitoksellanne

on kytss ja mik on niiden lukumr?
Mitattu muuttuja Mittaus- Mittausten sijainti Kuinka montaa Toimivatko

/ kytetty pisteiden mittauksista mittalaitteet
jatkuva-toiminen luku- kytetn kunnolla?
mittalaite mr on-line-
ohjaukseen?
Kyll
Ei
Liuennut happi Kyll
(DO) Ei
Kyll
Ei
Kyll
Ei
Kyll
Ammonium
Ei
Kyll
Ei
Kyll
Ei
Nitraatti
Kyll
Ei
Kyll
Ei
Fosfaatti
Kyll
Ei
Kyll
Ei
Kyll
Kiintoaine
Ei
Kyll
Ei
____________________________________________________________________________________________

Mitattu muuttuja / Mittauspisteiden Kuinka montaa Toimivatko
kytetty jatkuva- lukumr mittauksista mittalaitteet
toiminen mittalaite kytetn kunnolla?
on-line-
ohjaukseen?
Tulovirtaama Kyll
Ei
Veden pinnankorkeus Kyll
Ei
Ilmavirtaama Kyll
ilmastuksessa Ei
Ilmanpaine Kyll
ilmastuksessa Ei
Sameus Kyll
Ei
Lietepatjan korkeus Kyll
Ei
UV-absorptio Kyll
Ei
COD Kyll
Ei
BOD Kyll
Ei
Respirometri Kyll
Ei
pH Kyll
Ei
Redox-potentiaali Kyll
Ei
Johtokyky Kyll
Ei
Metaani Kyll
Ei
Lmptila Kyll
Ei
____________________________________________________________________________________________

Laitoksellanne kytss olevia mittauksia, mittalaitteita tai automaattisia
analysaattoreita, joita ei ole ylpuolella olevassa listassa:
Mitattu muuttuja / Kuinka montaa

Toimivatko
kytetty jatkuva- mittauksista
Mittauspisteiden lukumr mittalaitteet
toiminen kytetn
kunnolla?
mittalaite on-line-ohjaukseen?
Kyll
Ei
Kyll
Ei
Kyll
Ei
Kyll
Ei
2A. Kuinka aktiivilietteen ilmastus on ohjattu laitoksellanne?
Sama vakioilmamr koko vuorokauden ajan
Ilmastuksen ohjaus perustuu etukteen mriteltyihin aikoihin
Ilmastuksen ohjaus perustuu yhteen liuenneen hapen mittaukseen
Liuenneen hapen profiilin ohjaus
Ts. ohjaus perustuu ilmastusaltaan ilmastetun osan jakamiseen useisiin lohkoihin ja

liuenneen hapen mittaukseen kytetn useaa mittalaitetta
Ilmastuksen ohjaus perustuu ammonium-mittaukseen
Jokin muu ilmastuksen ohjausmenetelm, mik?
____________________________________________________________________________________________

3A. Listaus laitoksen trkeimmist ohjauksista
Toimiiko
Onko ohjauksen
Ohjaus Ohjauksen laatu ohjaus
vaihtelualue sopiva?
kunnolla?
Tulopumppaus On/off-ohjaus Alue on sopiva Kyll
Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Palautuslietteen On/off-ohjaus Alue on sopiva Kyll
pumppaus Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Ylijmlietteen On/off-ohjaus Alue on sopiva Kyll
pumppaus Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Ilmastus, On/off-ohjaus Alue on sopiva Kyll
stventtiilit Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Ilmastus, On/off-ohjaus Alue on sopiva Kyll
kompressorit Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Saostuskemikaalin On/off-ohjaus Alue on sopiva Kyll
(esim. ferro) sytt Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
____________________________________________________________________________________________

Toimiiko
Onko ohjauksen
Ohjaus Ohjauksen laatu ohjaus
vaihtelualue sopiva?
kunnolla?
Alkalointikemikaalin On/off-ohjaus Alue on sopiva Kyll
sytt Jatkuva ohjaus Alue on liian laaja Ei
Alue on liian kapea
Vaihtelualue on vr
Muuta, mit? On/off-ohjaus Alue on sopiva Kyll
Alue on liian kapea
Vaihtelualue on vr
Muuta, mit? On/off-ohjaus Alue on sopiva Kyll
Alue on liian kapea
Vaihtelualue on vr
4A. Kuinka monta kehittynytt sdint laitoksellanne on kytss?
Kokonaismr:
Minklaisia kehittyneit stimi laitoksellanne on ja miss ohjauksessa niit kytetn?
Kehittyneill stimill tarkoitetaan PID-sdint kehittyneemp menetelm,

esim. adaptiivista st, sumeaa st tai malleihin perustuvaa st.
5A. Onko laitoksellanne kytss automaatiota, joka reagoi poikkeuksellisiin tilanteisiin

tai joilla ennustetaan sellaisia? Minklaisia?
6A. Onko laitoksellanne kytetty prosessimallinnusta?
Kyll Ei
Onko laitoksellanne omaa mallinnusohjelmaa?

____________________________________________________________________________________________

Kyll Mik ohjelma?
Ei
Kytttek mallinnusta apuna laitoksenne ohjauksessa?
Kyll Ei
Mihin muihin tarkoituksiin olette kyttneet mallinnusta?

Mikli laitoksellanne ei toistaiseksi ole kytetty mallinnusta, oletteko harkinneet sen kytt?
Kyll, mallinnusta
Kyll, mallinnusta ja oman ohjelman hankintaa
Ei
____________________________________________________________________________________________

Mielipiteenne
7A. Miss mielestnne on paras mahdollisuus parantaa laitoksenne suorituskyky liittyen

instrumentaatioon, ohjaukseen ja automaatioon?
8A. Minklaisena arvioisitte nykyisen instrumentaation ja ohjausjrjestelmn kytn

laitoksellanne?
Niist olisi mahdollista hyty paljon enemmn
Niist olisi mahdollista hyty enemmn
Niist olisi luultavasti mahdollista hyty enemmn
Niist ei pystyt hytymn enemp
9A. Mink arvioitte trkeimmksi pullonkaulaksi laitoksenne toiminnan parantamisessa?
10A. Mink arvioitte suurimmaksi uhaksi tulevaisuudessa laitoksenne toiminnalle?
11A. Mit mielt olette tst vittmst? Instrumentaatio, ohjaus ja automaatio tulevat
olemaan jtevedenpuhdistamoilla entist trkempi lhitulevaisuudessa.
Tysin samaa mielt
Jokseenkin samaa mielt
En osaa sanoa
Jokseenkin eri mielt
Tysin eri mielt
12A. Onko laitoksellenne asennettu uutta ohjaus- tai sttekniikkaa viimeisen viiden
vuoden aikana?
Kyll Ei
____________________________________________________________________________________________

Mikli vastaus edelliseen kysymykseen on mynteinen, mit parannuksia laitoksellanne
on saavutettu uuden ohjaus- tai sttekniikan ansiosta?
Kommentteja tst kyselylomakkeesta tai siin olevista kysymyksist:
____________________________________________________________________________________________

KYSELY JTEVEDENPUHDISTAMOIDEN
KOKOONPANOSTA JA TOIMINNASTA
Teidn yhteystietonne
Nimi Tehtvnimike
Shkpostiosoite (Ty)puhelinnumero
Tietoja jtevedenpuhdistamosta
Laitoksen nimi
Osoite
Paikkakunta Postinumero
____________________________________________________________________________________________

Laitoksen tyyppi ja operointi
Shkisesti tytetyt kyselylomakkeet voi palauttaa liitetiedostoina osoitteeseen

henri.haimi@tkk.fi.
Jotta erilaisten puhdistamoiden ohjauksesta ja automaatiosta saataisiin edustava otos

johtoptsten tekemiseen, pyytisimme teit vastaamaan seuraaviin kysymyksiin koko vuoden
tilastotietojen pohjalta. Olkaa hyv ja valitkaa vuosi 2006 tai 2007 .
Mikli joihinkin kysymyksiin lytyy vastaukset lhettmistnne vuosiraporteista tai

kyselylomakkeen liitteen olevasta laitoksenne virtauskaaviosta*, voitte viitata niihin sen sijaan,
ett vastaatte kysymyksiin erikseen lomakkeeseen.
* Ari Kangas 2004. Jtevedenpuhdistamojen toiminta ja toteutukset,

Vesi- ja viemrilaitosyhdistyksen monistesarja Nro 15, 12 s.
1B. Mist vuodesta lhtien laitoksenne on ollut kytss?
Laitos on ollut kytss vuodesta
2B. Mikli laitostanne on saneerattu sen kyttn oton jlkeen, milloin se on viimeksi
tapahtunut? Mitk ovat viimeisimmt merkittvt saneeraustoimenpiteet, mahdolliset
automaation uudistukset sek syyt niihin?
Laitosta on viimeksi saneerattu vuonna
Viimeisimmt merkittvt saneeraustoimenpiteet:
Automaation uudistukset:
Syyt saneeraukseen:
3B. Kuinka suuri osuus laitoksenne nykyisest virtaamasta ja kuormasta (esim. BOD7 tai
N) on teollisuusjtevesi?
Teollisuusjteveden osuus % virtaamasta
Teollisuusjteveden osuus % kuormasta
____________________________________________________________________________________________

4B. Mik on jteveden keskimrinen lmptila?
Keskimrinen jteveden lmptila on C
Jteveden lmptila on yli 12 C kk vuodesta
Jteveden minimilmptila ko. vuotena oli C
5B. Kuinka suuren osan ajasta laitoksellanne on henkilkuntaa paikalla?
Viiten pivn viikossa ainoastaan pivisin
Seitsemn pivn viikossa ainoastaan pivisin
Seitsemn pivn viikossa sek isin ett pivisin
Miehittmtn laitos
Joku muu vaihtoehto tuntia viikossa
6B. Kuinka paljon henkilkuntaa laitoksellanne tyskentelee?
Tysiaikaisten tyntekijiden lukumr
Osa-aikaisten tyntekijiden lukumr
7B. Mik on laitoksenne vuosittainen shknkulutus?
Kokonaiskulutus kWh/vuosi
Mit ilmoittamaanne kokonaiskulutukseen on sisllytetty?
Mikli laitoksellanne on mitattu ksittelyprosessin eri osien shknkulutus, olkaa hyv ja

vastatkaa mys alla oleviin kysymyksiin.
Esiksittelyn kulutus kWh/vuosi

(Kokonaiskulutus ennen vesiprosessin biologista osaa)
Biologisen osan kulutus kWh/vuosi
Ilmastuksen osuus % biologisen osan kulutuksesta
Lietteen ksittelyn kulutus kWh/vuosi
Kuinka paljon shk tuotetaan? kWh/vuosi
____________________________________________________________________________________________

8B. Kuinka paljon laitoksellanne kytetn kemikaaleja vuosittain? (Saostuskemikaali,
alkalointikemikaali ja ulkopuolisena hiililhteen kytetty kemikaali)
Kyttmnne saostuskemikaali:
Saostuskemikaalin vuotuinen kulutus tonnia / vuosi
Saostuskemikaalin annostelupisteet:
Kyttmnne alkalointikemikaali:
Alkalointikemikaalin vuotuinen kulutus: tonnia / vuosi
Mikli laitoksellanne kytetn ulkopuolista hiililhdett (esim. metanoli) denitrifikaation ja/tai

biologisen fosforinpoiston tehostamiseen, vastatkaa mys kysymyksiin alla.
Kyttmnne ulkopuolinen hiililhde:
Vuosittainen kulutus tonnia / vuosi
Hiililhteen annostelupisteet:
9B. Biologisen prosessin ajo
Mik on biologisen prosessin keskimrinen lieteik?
Talvella d
Kesll d
Mik on biologisen prosessin pasiallinen ajotapa?
Prosessia ajetaan tavoitelietein mukaan
Prosessia ajetaan tavoitelietepitoisuuden mukaan ja lieteik lasketaan kontrolliksi
Jokin muu ajotapa, mik?
____________________________________________________________________________________________

Miten ilmastettujen lohkojen mr mrytyy? (Vastataan vain, jos laitoksella ajetaan
kokonaistypenpoistoa.)
Henkilkunta asettaa manuaalisesti
Automaattisesti kello-ohjauksella
Automaattisesti ammoniumtyppimittauksen perusteella
Jokin muu tapa, mik?
10B. Tyttyivtk laitoksenne lupaehdot kyseisen vuotena?
Kyll Ei
Mikli lupaehdot eivt ole tyttyneet, mink suureen osalta ja min ajanjaksona nin ei
ole tapahtunut?
11B. Kuinka usein jtevett ohitettiin kyseisen vuotena?
Pelkstn puhdistamon biologinen osa ohitettiin
mr vuoden aikana
Koko ksittelyprosessi ohitettiin
mr vuoden aikana
12B. Jos jtevedenpuhdistamollanne on jotain poikkeuksellisia piirteit, olkaa hyv ja

kuvailkaa niit.
Eptavallisen suuria vaihteluita ksiteltvn jteveden laadussa tai virtaamassa:
Poikkeuksellisia teollisuusjtevesi:
Eptavallista prosessitekniikkaa:
Poikkeuksellisia lupaehtoja:
Jotain muuta, mit?
____________________________________________________________________________________________

13B. Mink tyyppinen ksittelyprosessinne on typenpoiston osalta?
Esidenitrifikaatio
Samanaikainen nitrifikaatio / denitrifikaatio
Jlkidenitrifikaatio
Vaiheittainen nitrifikaatio / denitrifikaatio
Pelkk nitrifikaatio
Joku muu, olkaa hyv ja tarkentakaa:
14B. Kuinka suurta virtaamaa laitoksenne on suunniteltu ksittelemn?
Keskimrinen mitoitusvirtaama m3/vrk
Maksimimitoitusvirtaama m3/tunti
15B. Kuinka suuri on laitoksenne nykyinen toteutunut virtaama?
Nykyinen keskivirtaama m3/vrk
16B. Kyselylomakkeen liitteen on laitoksenne prosessikaavio allastilavuuksineen ja

kemikaalien annostelupisteineen vuodelta 2003. Onko prosessinne edelleen samanlainen
ja reaktoritilavuudet samat?
Kyll
Ei Mit muutoksia nykyisess prosessissanne on liitteen olevaan prosessikaavioon

verrattuna?
____________________________________________________________________________________________

17B. Kuinka suuria biologisten reaktoreiden (sislten aerobisen, anoksisen ja
anaerobisen tilavuuden), selkeytysyksikiden sek mahdollisen tasausaltaan tilavuudet
ja ksittelylinjojen lukumrt ovat? Jos laitoksellanne on kytss jokin
jlkiksittely, olkaa hyv ja kuvailkaa kyseist ksittelyprosessia.
Aktiiviliete linjaa m3
Esiselkeytys linjaa m3
Jlkiselkeytys linjaa m3
Tasausallas m3
Jlkiksittely, mik?
18B. Mit aineita laitoksen on suunniteltu poistavan orgaanisen aineen ja fosforin

lisksi? (Rastittakaa kaikki sopivat vaihtoehdot)
Ammoniumtyppi (mutta ei nitraattityppe)
Kokonaistyppi (sek ammonium ett nitraattityppi)
Fosfori (biologinen ksittely)
19B. Kuinka paljon laitoksellanne syntyy lietett?
Kokonaispaino tonnia / vuosi
Lietteen kuiva-ainepitoisuus %
____________________________________________________________________________________________

Appendix B. Key figures of the WWTP survey
Median Average Min Max
Design flow rate, m3/d 21 500 38 300 2 500 260 000
3
Max. Design flow rate, m /h 1 550 2 980 630 25 000
3
Current flow rate, m /d 13 250 29 200 2 150 260 000
Current flow rate / design flow rate, % 68 69 35 105
Sludge age during winter, total N or NH4-N removal, d 14.5 15.0 6.0 30.0
Sludge age during summer, total N or NH 4-N removal, d 9.5 10.8 5.0 20.0
Sludge age during winter, only organic matter removal, d 5.0 8.6 3.0 20.0
Sludge age during summer, only organic matter removal, d 3.5 5.4 2.0 10.0
Average temperature of wastewater, C 12.3 12.3 8.7 16.0
Min. temperature of wastewater, C 7.0 6.8 3.3 10.1
Temperature of wastewater above 12C, months per year 6.0 6.2 1.5 11.0
Share of industrial wastewater of the flow rate, % 10.0 10.5 0 30.0
Share of industrial wastewater of the load, % 15.5 19.8 0 60.0
Influent COD / total N 11.0 11.9 8.9 18.4
Sludge production, kg TS/kg BOD7 1.10 1.19 0.54 2.48
Dry solids content of sludge, % 23.6 23.5 6.0 32.0
3
Energy consumption / influent flow rate, kWh/m 0.47 0.51 0.17 1.00
Set-point of dissolved oxygen concentration in aeration, mg/l 2.5 2.6 2.0 3.1
Number of full-time employees 6 9.8 2 50
____________________________________________________________________________________________


Ica Report Haimi Et Al PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ica Report Haimi Et Al PDF

Uploaded by

Copyright:

Available Formats

Teknillinen korkeakoulu Vesihuoltotekniikan julkaisu

Helsinki University of Technology Water and Wastewater Engineering

ADVANCED OPERATION AND CONTROL METHODS OF MUNICIPAL

Henri Haimi, Michela Mulas, Kristian Sahlstedt and Riku Vahala

ADVANCED OPERATION AND CONTROL METHODS OF MUNICIPAL

Henri Haimi, Michela Mulas, Kristian Sahlstedt and Riku Vahala

Helsinki University of Technology

Advanced operation and control methods of municipal 2

Advanced operation and control methods of municipal 3

Advanced operation and control methods of municipal 4

Advanced operation and control methods of municipal 5

1.3 Report overview

Advanced operation and control methods of municipal 6

Advanced operation and control methods of municipal 7

Advanced operation and control methods of municipal 8

2.1 Activated sludge process models

Advanced operation and control methods of municipal 9

Secondary settler models

Advanced operation and control methods of municipal 10

ASP Simplified models

Advanced operation and control methods of municipal 11

Before modelling a preliminary sampling campaign on influent was conducted to investigate

Advanced operation and control methods of municipal 12

Rodrguez et al. (2009) presented a wastewater treatment modelling methodology based on

Advanced operation and control methods of municipal 13

Advanced operation and control methods of municipal 14

Advanced operation and control methods of municipal 15

Advanced operation and control methods of municipal 16

2.4 Process Simulation Environment

Specific commercial environments usually contain extended libraries of predefined process

Advanced operation and control methods of municipal 17

Advanced operation and control methods of municipal 18

Advanced operation and control methods of municipal 19

3.1 Feedback control algorithms

Figure 4. Typical components of a feedback control loop.

Advanced operation and control methods of municipal 20

Advanced operation and control methods of municipal 21

3.2 Feedforward Control

Advanced operation and control methods of municipal 22

In practical application, feedforward control is normally used in combination with feedback

3.3 Advanced control algorithms

Advanced operation and control methods of municipal 23

Figure 7. Configuration of cascade control loops in Cho et al. (2002).

Model predictive control

Figure 8. Basic structure of MPC (Camacho and Bordons, 2004).

Advanced operation and control methods of municipal 24

Advanced operation and control methods of municipal 25

Fuzzy logic control

Advanced operation and control methods of municipal 26

IF such and such states of process variables (i.e. inputs) are

Advanced operation and control methods of municipal 27

Artificial neural networks

Advanced operation and control methods of municipal 28

3.4 Control applications

Advanced operation and control methods of municipal 29

Advanced operation and control methods of municipal 30

Advanced operation and control methods of municipal 31

Advanced operation and control methods of municipal 32

Multivariable statistical techniques were used in analysis of plant-wide WWTP control

Advanced operation and control methods of municipal 33

Real plant control examples

A hierarchical two-level control to track DO reference trajectory in activated sludge process