Mechanistic modelling of pulp and paper mill wastewater treatment plants

Similar documents
Modelling of Wastewater Treatment Plants

A General Description of the IAWQ Activated Sludge Model No. 1

A general Description of the IAWQ activated sludge model No.1

CSR Process Simulations Can Help Municipalities Meet Stringent Nutrient Removal Requirements

EnviTreat Laboratory Examples of Studies

Examples of Studies conducted by

Combined Optimization of the Biological Nitrogen Removal in Activated Sludge-Biofilm Wastewater Treatment Systems

NEW BIOLOGICAL PHOSPHORUS REMOVAL CONCEPT SUCCESSFULLY APPLIED IN A T-DITCH PROCESS WASTEWATER TREATMENT PLANT

ENHANCING THE PERFORMANCE OF OXIDATION DITCHES. Larry W. Moore, Ph.D., P.E., DEE Professor of Environmental Engineering The University of Memphis

USING NUMERICAL SIMULATION SOFTWARE FOR IMPROVING WASTEWATER TREATMENT EFFICIENCY

Contents General Information Abbreviations and Acronyms Chapter 1 Wastewater Treatment and the Development of Activated Sludge

HOW TO SELECT SRT. SRT is selected based on the following treatment requirements/objectives:

SIMPLE and FLEXIBLE ENERGY SAVINGS And PERFORMANCE ENHANCEMENT for OXIDATION DITCH UPGRADES

operation of continuous and batch reactors. Contrary to what happens in the batch reactor, the substrate (BOD) of the wastewater in the continuous rea

ADVANCING NOVEL PROCESSES FOR BIOLOGICAL NUTRIENT REMOVAL

A General Description of the Activated Sludge Model No. 1 (ASM1)

Enhancing activated sludge nitrification through seeding from a downstream nitrifying fixed-film reactor

IWA Publishing 2012 Water Practice & Technology Vol 7 No 3 doi: /wpt

CRUDE COD CHARACTERISTICS SIGNIFICANT FOR BIOLOGICAL P REMOVAL: A U.K. EXAMPLE

TWO YEARS OF BIOLOGICAL PHOSPHORUS REMOVAL WITH AN ADVANCED MSBR SYSTEM AT THE SHENZHEN YANTIAN WASTEWATER TREATMENT PLANT

CTB3365x Introduction to Water Treatment

Probabilistic Modeling of Two-Stage Biological Nitrogen Removal Process: Formulation of Control Strategy for Enhanced Process Certainty

AMPC Wastewater Management Fact Sheet Series Page 1

AMPC Wastewater Management Fact Sheet Series Page 1

Effect of the start-up length on the biological nutrient removal process

New Developments in BioWin 4.0

Evaluation of an ASM1 model calibration procedure on a municipal industrial wastewater treatment plant

modelling to support re-design and operation of full-scale

Removal of High C and N Contents in Synthetic Wastewater Using Internal Circulation of Anaerobic and Anoxic/Oxic Activated Sludge Processes

Mathematical Modelling of Wastewater Treatment Plant of Žiar nad Hronom

Municipal Wastewater Treatment Improvement Using Computer Simulating

NITROGEN REMOVAL USING TERTIARY FILTRATION. Suzie Hatch & Colum Kearney. Sydney Water Corporation

General Information on Nitrogen

SIMULATION AND CALIBRATION OF A FULL- SCALE SEQUENCING BATCH REACTOR FOR WASTEWATER TREATMENT

Characterisation and biological treatability of Izmit industrial and domestic wastewater treatment plant wastewaters

ENHANCED BIOLOGICAL PHOSPHORUS REMOVAL WITHIN MEMBRANE BIOREACTORS. 255 Consumers Road Toronto, ON, Canada, M2J 5B6

Wastewater Pollutants & Treatment Processes. Dr. Deniz AKGÜL Marmara University Department of Environmental Engineering

Module 11 : Water Quality And Estimation Of Organic Content. Lecture 14 : Water Quality And Estimation Of Organic Content

MODELING OF AN IFAS PROCESS WITH FUNGAL BIOMASS TREATING PHARMACEUTICAL WASTEWATER Main Street West Hamilton, ON, Canada L8S 1G5

Wastewater Terms for Permit Applications

BIOLOGICAL PHOSPHOROUS REMOVAL AN OPERATOR S GUIDE

Evaluation of Different Nitrogen Control Strategies for a Combined Pre- and Post-Denitrification Plant

Advances in Multiobjective Optimization In Practice

WASTEWATER TREATMENT

Overview of Supplemental Carbon Sources for Denitrification and Enhanced Biological Phosphorus Removal

Dynamics of Wastewater Treatment Systems

Department of Civil Engineering-I.I.T. Delhi CVL723 Problem Set_2_Feb6_15

ANOXIC BIOREACTOR SIZING

1/11/2016. Types and Characteristics of Microorganisms. Topic VI: Biological Treatment Processes. Learning Objectives:

SBR PROCESS FOR WASTEWATER TREATMENT

BENCH SCALE DENITRIFICATION STUDY. for. Sucrosolutions TM for Water Sugar Australia

Dynamic Simulation of Petrochemical Wastewater Treatment Using Wastewater Plant Simulation Software

Environmental Biotechnology Cooperative Research Centre Date submitted: March 2008 Date published: March 2011

Increasing Denitrification in Sequencing Batch Reactors with Continuous Influent Feed

PROCESS SIMULATOR FOR WASTEWATER TREATMENT PLANT

PHENOLIC WASTEWATER TREATMENT

Proceedings of the World Congress on Engineering 2018 Vol I WCE 2018, July 4-6, 2018, London, U.K.

Kinetic optimization of the activated sludge denitrification process

COMPARISON OF SBR AND CONTINUOUS FLOW ACTIVATED SLUDGE FOR NUTRIENT REMOVAL

A practical and sound model calibration procedure applied to the WWTP of Eindhoven

PLANNING FOR NUTRIENT REMOVAL: WHAT STEPS CAN WE BE TAKING NOW?

WASTEWATER TREATMENT PLANT OPTIMIZATION USING A DYNAMIC MODEL APPROACH

with sewage effluent Nutrient enrichment ( Eutrophication ) Algal Blooms Deaeration of the watercourse oxidation of ammonia a potable.

Choices to Address Filamentous Growth

336098: DYNAMIC MODELLING AND SIMULATION OF ANAEROBIC DIGESTER FOR HIGH ORGANIC STRENGTH WASTE

WASTEWATER DEPARTMENT. Bentonville Wastewater Treatment Plant Facts:

Refinery Wastewater Process Modeling with GPS-X

The Activated Sludge Model No 1 ASM1. Bengt Carlsson

AquaNereda Aerobic Granular Sludge Technology

Chapter 4: Advanced Wastewater Treatment for Phosphorous Removal

Supplemental Information for

/ Marley MARPAK Modular Biomedia /

Preparing for Nutrient Removal at Your Treatment Plant

FIRST APPLICATION OF THE BABE PROCESS AT S-HERTOGENBOSCH WWTP. Pettelaarpark 70, PO Box GA, s-hertogenbosch, The Netherlands

Waste water treatment

RTC-N: Modelling Ammonia-Based Aeration Control in Real Time for Deeper Process Control

Chapter 9 Nitrification

BIOLOGICAL PHOSPHORUS AND NITOGEN REMOVAL IN A SQUENCHING BATCH MOVING BED BIOFILM REACTOR

Water and Wastewater Engineering Dr. Ligy Philip Department of Civil Engineering Indian Institute of Technology, Madras

ENVE 302 Environmental Engineering Unit Processes DENITRIFICATION

Feedforward aeration control of a Biocos wastewater treatment plant

TIES598 Nonlinear Multiobjective Optimization Applications spring 2017

Optimizing Nutrient Removal. PNCWA - Southeast Idaho Operators Section Pocatello, ID February 11, 2016 Jim Goodley, P.E.

Sewage Treatment Plant Upgrade Applying Wastewater Process Simulation

Application of the AGF (Anoxic Gas Flotation) Process

Aeration Basics the Bug s Eye View

A Roadmap for Smarter Nutrient Management in a Carbon and Energy Constrained World. Samuel Jeyanayagam, PhD, PE, BCEE

A Study on Sludge Reduction in Sewage using Microbial Catalysts

Secondary Treatment Process Control

AquaPASS. Aqua MixAir System. Phase Separator. System Features and Advantages. Anaerobic. Staged Aeration. Pre-Anoxic.

Biological Phosphorous Removal Is Coming! Michigan Water Environment Association Annual Conference, June 23, 2008; Boyne Falls MI

Bio-Tiger Model Biokinetic Model that Simulates Activated Sludge Processes

Water Technologies. The AGAR Process: Make Your Plant Bigger Without Making it Bigger

Nutrient Removal Processes MARK GEHRING TECHNICAL SALES MGR., BIOLOGICAL TREATMENT

RE ENGINEERING O&M PRACTICES TO GET NITROGEN & PHOSPHORUS REMOVAL WITHOUT FACILITY UPGRADES

DEGRADATION OF AMMONIA IN AN INTEGRATED ANOXIC-AEROBIC CLARIFIER SYSTEM

Energy Reduction and Nutrient Removal in WWTPs Using Feed- Forward Process Control

Key Points. The Importance of MCRT/SRT for Activated Sludge Control. Other (Confusing) Definitions. Definitions of SRT

Secondary Wastewater Treatment

ANAEROBIC TREATMENT OF PAPER MILL WASTEWATER

Transcription:

CONTROL ENGINEERING LABORATORY Mechanistic modelling of pulp and paper mill wastewater treatment plants Jukka Keskitalo and Kauko Leiviskä Report A No 41, January 2010

University of Oulu Control Engineering Laboratory Report A No 41, January 2010 Mechanistic modelling of pulp and paper mill wastewater treatment plants Jukka Keskitalo and Kauko Leiviskä University of Oulu, Control Engineering Laboratory Abstract: This report provides a brief review on characteristics of pulp and paper wastewaters, activated sludge processes for wastewater treatment and the state of the art Activated Sludge Models (ASM). Literature on calibration of the ASMs and on application examples is reviewed more thoroughly. Additionally, a case study on mechanistic modelling and model calibration of pulp and paper industry wastewater treatment plant is presented. The widely used ASMs have been developed mainly for modelling the treatment of municipal wastewater. Mechanistic modelling of activated sludge treatment of pulp and paper mill wastewater requires some special considerations, as the wastewater is nutrient deficient. Calibration of the ASMs remains the weakest link in activated sludge modelling, as the models are complex and usually overparameterised to a given problem. In the case study, results of a measurement campaign from a pulp mill wastewater treatment plant are presented. Oxygen uptake rate (OUR) measurements and conventional wastewater analyses were made with sludge and wastewater sampled from the treatment plant. The results were utilised in calibrating a modified ASM no. 1 for the treatment plant and for wastewater characterisation. The model performance was validated by running a simulation with ten months of influent process data from the mill databases as inputs to the model. Model predictions of effluent quality were then compared to measured values. The measured and simulated values were in good agreement for most of the simulation period. This report is an extension to journal article by Keskitalo et al. [1]. This report provides more background information and sensitivity analysis of the model to better justify the calibration procedure. Keywords: pulp and paper wastewater, model calibration, modified activated sludge model no. 1, computer simulation, full-scale WWTP ISBN 978-951-42-6110-7 ISSN 1238-9390 University of Oulu Control Engineering Laboratory P.O. Box 4300 FIN-90014 University of Oulu

1 Introduction... 1 2. Review... 2 2.1 Pulp and paper mill effluents... 2 2.2 Activated sludge process... 3 2.3 Activated Sludge Models... 5 2.4 Calibration of the Activated Sludge Models... 6 2.4.1 Systematic experimental and experience-based ASM calibration protocols... 6 2.4.2 Systematic protocols for assessing parameter identifiability and calibration of parameters of the ASMs... 7 2.4.3 Examples on ASM calibration... 9 2.4.4 Discussion on the different approaches to ASM calibration... 12 3. Case study on activated sludge modelling... 13 3.1 Materials & Methods... 13 3.1.1 Plant & data description... 13 3.1.2 Analytical work... 14 3.1.3 Model structure... 16 3.1.4 Sensitivity analysis... 18 3.1.5 Wastewater characterisation... 20 3.1.6 Calibration procedure... 22 3.2 Results & Discussion... 23 3.2.1 OUR measurements and wastewater analyses... 23 3.2.2 Sensitivity analysis... 26 3.2.3 Wastewater characterisation... 27 3.2.4 Calibration... 29 3.2.5 Simulation... 30 4. Conclusions... 35 5. References... 36

SYMBOLS AND ABBREVIATIONS AOX Adsorbable organically bound halogen ASM Activated Sludge Model ASM1 Activated Sludge Model No. 1 ASM2 Activated Sludge Model No. 2 ASM2d Activated Sludge Model No. 2d ASM3 Activated Sludge Model No. 3 Bio-P Biological phosphorus removal BOD Biological oxygen demand COD Chemical oxygen demand CSTR Continuous stirred-tank reactor DO Dissolved oxygen FIM Fischer information matrix IWA International Water Association MLSS Mixed liquor suspended solids N Nitrogen OUR Oxygen uptake rate P Phosphorus PAO Phosphorus accumulating organism SA Sensitivity analysis S I Soluble inert material S ND Soluble biodegradable organic nitrogen S NH Ammonium nitrogen S NI Inert soluble nitrogen S NO Nitrate nitrogen SOUR Specific oxygen uptake rate S P Soluble biodegradable phosphorus SS Suspended solids S S Readily biodegradable substrate X I Particulate inert material X NB Active mass nitrogen X NI Inert particulate nitrogen X NP Nitrogen in products arising from biomass decay X PB,1 Minimum phosphorus in active biomass X PB,2 Additional phosphorus in active biomass X PD Particulate organically bound phosphorus X PP Phosphorus in products arising from biomass decay X S Slowly biodegradable substrate VSS Volatile suspended solids WWTP Wastewater treatment plant Y H Heterotrophic yield coefficient δ msqr Local sensitivity measure

1 1 INTRODUCTION Modelling activated sludge wastewater treatment has been extensively studied since publication of the first Activated Sludge Model, the ASM1, in 1987. Modelling has become an accepted practice in treatment plant design, teaching and research. Even though the Activated Sludge Models were introduced for modelling municipal wastewater treatment, the models have been successfully applied for modelling the treatment of some industrial wastewaters. This report attempts to extend activated sludge modelling and model calibration to pulp and paper wastewaters, which has not yet received much attention. The report begins with an introduction to the characteristics of pulp and paper wastewaters, activated sludge process, the state of the art activated sludge models and model calibration. The state of the art models are complex and usually overparameterised to a given problem [2]. Calibration of the models remains the weakest link in activated sludge modelling [3]. A case study of modelling pulp and paper industry wastewater treatment plant with a modified ASM no. 1 is presented. The model is calibrated with a simplified calibration protocol utilising results from oxygen uptake rate (OUR) measurements. The model is used for simulating the full-scale pulp mill wastewater treatment plant. Simulation results are compared to measured process data. This report is based on the same process data and experimental results as [1]. The report complements the article by providing more background information and application of sensitivity analysis of the model to better justify the calibration procedure.

2 2. REVIEW 2.1 Pulp and paper mill effluents Even though the pulp and paper making industry remains a major consumer of freshwater globally, its emissions to waterways have been significantly reduced. For example, BOD 7 emissions from the industry in Finland have been reduced by about 90% since the 1970 while the production of paper and board has become threefold and production of pulp has doubled. Emissions of adsorbable organically bound halogens have also been decreased to a fraction of their past levels. Similar trend, however, has not been seen for nutrients nitrogen and phosphorus. [4] Challenges still remain in reducing incidental discharges, which have become a significant fraction of the total discharges. Disturbances in upstream production processes cause changes in wastewater quality and volume, which may further disturb the external biological wastewater treatment and cause violations of effluent quality limits. [5] Effluents from pulp and paper mills contain wood components, chemicals used in the processes and their reaction products, fillers and auxiliary chemicals. Effluent compositions vary considerably and only part of the effluent components have been identified so far. Most of the contaminants are in solid form but some are colloidal or dissolved. Effluents from chemical pulping are highly coloured due to the dissolved lignin. Untreated effluents contain components such as fatty acids and resin acids, which are toxic to aquatic life. Levels of nitrogen and phosphorus in pulp and paper mill effluents are usually low compared to municipal sewage. [4] The different components in wastewater are lumped together in commonly used measurements. The most common measurements are biological oxygen demand (BOD), chemical oxygen demand (COD), suspended solids (SS), phosphorus (P), nitrogen (N), adsorbable organically bound halogen (AOX), chloro-organics, colour and toxicity [4]. These measurements are used in design and operation of wastewater treatment plants and in defining water discharge limits. As for the individual compounds, the measured characteristics of wastewater are different depending on the type of the mill. Typical wastewater loads from several types of mills before external treatment are given in Table 1.

3 Table 1. Typical wastewater loads from different types of mills before external treatment [4]. Effluent volume [m 3 /t] COD [kg/t] BOD [kg/t] Suspended solids [kg/t] Nitrogen [g/t] Phosphorus [g/t] Sulphate pulp, 20-60 20-30 5-10 12-15 200-400 80 unbleached Sulphate pulp, 60-100 60-120 18-25 12-18 300-500 120 conventional bleaching Sulphite pulp, 150-200 60-100 30-40 20-40 100-200 60 conventional bleaching Thermomechanical 6-15 40-80 15-25 10-30 100-200 70 pulp, unbleached Thermomechanical 6-15 60-100 20-40 10-30 200-300 100 pulp, peroxide bleached Fine paper, coated 30-50 10-20 3-8 10-20 50-100 5 Newsprint 10-25 2-4 1-3 5-10 10-20 5 Tissue 20-40 3-6 1-3 5-10 50-80 8 2.2 Activated sludge process The external treatment of pulp and paper mill wastewaters begins commonly with primary treatment stages including solids removal, neutralisation, cooling and equalisation [6]. Possible processes for solids removal are sedimentation and flotation. High solids removal over 80% is achieved in primary treatment. However, the insufficient removal of organic material requires secondary treatment. [7] The most commonly applied secondary treatment methods for pulp and paper effluents are biological treatment in activated sludge process or aerated lagoon. The use of aerated lagoons has recently declined due their lower treatment performance. [6] Activated sludge process is available in numerous configurations from simple designs for organic material removal to more complex configurations, which are required for biological nutrient removal [8]. Emissions to waterways from pulp and paper industries are dominated by organic material [6]. Therefore simple process designs without biological nutrient removal are common in the industry. The basic activated sludge process consists of an aerated reactor where the sludge is kept in suspension, sedimentation tank for liquid-solids separation and recycle stream for returning separated solids back to the reactor [8]. The process is illustrated in Figure 1. Microorganisms which perform the treatment are kept in the suspension along with other solid material from the influent wastewater. The microorganisms oxidise dissolved and particulate carbonaceous organic material found in wastewater to produce energy and new cells. Nitrogen and phosphorus nutrients are needed for cell growth. [8]

4 Primary clarifier Aeration basin Secondary clarifier Influent Effluent Primary sludge Return activated sludge Biological sludge Figure 1. Complete mix activated sludge process. All processes for biological nitrogen removal require an aerobic zone for nitrification and an anoxic zone to complete the nitrogen removal by denitrification. The processes are grouped according to the location of the anoxic zone to preanoxic, postanoxic and simultaneous nitrification-denitrification processes. The preanoxic process is used most often. In the preanoxic process, nitrate is fed to the anoxic reactor by internal recycle from the aerobic reactor and by return activated sludge from the secondary clarifier. Electron donor in nitrification is biodegradable COD in the influent wastewater. The preanoxic process is illustrated in Figure 2. In the postanoxic process, nitrate is reduced in the anoxic reactor where very little biodegradable COD remains in the nitrified effluent. Denitrification can be achieved by endogenous respiration if retention time is sufficiently long, or by addition of external carbon source such as methanol. [8] Primary clarifier Internal recycle Secondary clarifier Influent Anoxic Aerobic Effluent Primary sludge Return activated sludge Biological sludge Figure 2. Preanoxic process for biological nitrogen removal. Biological phosphorus removal (Bio-P) can be accomplished by an anaerobic zone followed by an aerobic zone. Phosphorus removal is initiated in the anaerobic zone where phosphorus accumulating organisms (PAOs) convert volatile fatty acids to internal carbon storage products with energy obtained from releasing stored polyphosphates. Under aerobic conditions where external biodegradable substrate is less available PAOs utilise the storage products and take up polyphosphates. Phosphorus removal is achieved by wasting the phosphorus rich sludge from aerobic zone. Bio-P can be combined with the various biological nitrogen removal mechanisms. [8]

5 2.3 Activated Sludge Models The most widely used model for describing biological wastewater treatment processes is the International Water Association (IWA) (formerly IAWQ and IAWPRC) Activated Sludge Model No. 1 (ASM1) published in 1987. The goal in the model development was to develop the simplest model capable of describing carbon oxidisation, nitrification and denitrification accurately. [9] The ASM3 was also developed to describe biological nitrogen removal. The main differences between the two models are the role of storage polymers and the change of growth-decay-growth model to growth-endogenous respiration model in the ASM3. The changes were made to correct defects which were recognised in the usage of the ASM1 and to improve parameter identifiability. The ASM2 was developed to include description of biological phosphorus removal in the ASM1. The description of biological phosphorus removal was not complete in ASM2. Therefore the model was improved in ASM2d. [10] In the ASM1, the carbonaceous material is divided into readily biodegradable substrate (S S ), slowly biodegradable substrate (X S ), soluble inert material (S I ) and particulate inert material (X I ), heterotrophic biomass (X B,H ), autotrophic biomass (X B,A ) and inert particulate products arising from biomass decay (X P ). The total nitrogen is also divided into different state variables. It is hypothesised that the S S is directly oxidised by microorganisms while the X S has to first undergo hydrolysis. The decay of microorganisms produces X P, but also X S according to the death-regeneration hypothesis. [11] The process rates are modelled using Monod or first-order kinetics. [9] The model contains a total of eight processes with 19 parameters affecting 13 state variables. ASM2 is more complex than the ASM1 and has additional components and processes to describe biological phosphorus removal. In addition to heterotrophic and nitrifying organisms ASM2 has phosphorus-accumulating organisms. ASM2 also describes chemical precipitation of phosphorus. High molecular weight compounds must undergo hydrolysis before being utilised by microorganisms as in ASM1. However, in ASM2 hydrolysis processes depend on the electron acceptor conditions. [11] Growth-decaygrowth model of ASM1 was not changed until ASM3. ASM2d has an important extension to ASM2: denitrifying phosphorus accumulating organisms [11]. The ASM2 contains 19 processes affecting 19 state variables and the ASM2d contains 21 processes affecting 19 state variables. In ASM3 the decay process of ASM1 is replaced with the concept of endogenous respiration. The flows of COD in the growth and decay of heterotrophic and nitrifying bacteria are clearly separated in ASM3. As the processes are less interrelated, identifiability of the model should be better. ASM3 includes cell internal storage compounds and associated storage processes. All substrate must be stored before being utilised. [11] ASM3 with description of biological phosphorus removal was published in [12]. The ASM3 contains 12 processes affecting 13 state variables. The ASM3 with bio-p contains 23 processes with 32 parameters affecting 17 state variables.

6 The full description of the original Activated Sludge Models ASM1, ASM2, ASM2d and ASM3 including model equations and default parameters can be found in [11]. Numerous modifications to the original ASMs have been published. One example of a modified ASM is the Lindblom s model for nutrient deficient wastewaters [13]. Lindblom s model is described in more detail in Chapter 3.1.3 and is used in this work for modelling a fullscale pulp mill wastewater treatment plant. 2.4 Calibration of the Activated Sludge Models The parameter sets of the ASMs are not universal as most applications require adjusting the model parameters according to the characteristics of the treatment plant [2]. It has been even stated that calibration of the models is strictly required prior to application [3]. Calibration of the ASMs has traditionally been based on ad-hoc approaches and expert knowledge of the modeller. An overview of the methodologies that have been applied in calibration of the ASMs can be found in [14]. Systematic experience based protocols have been proposed lately in attempt to create a standard protocol for calibration of the ASMs [15-18]. There have also been attempts to automate the calibration procedure by developing approaches based on systems analysis [19-21]. These approaches address the problem of poor identifiability of the ASMs by analysing the practical identifiability of the models to find identifiable subsets of parameters. Optimisation algorithms can be then applied to tune the parameters of the identifiable subsets. Despite all the effort and numerous publications on calibrating the ASMs, none of the proposed calibration protocols has established status as the standard calibration protocol. 2.4.1 Systematic experimental and experience-based ASM calibration protocols Systematic protocols have been proposed for calibrating the ASMs, which remains the weakest link in modelling the activated sludge treatment [3]. Four such protocols are documented in the literature: The Dutch Foundation of Applied Water Research (STOWA) calibration protocol [15], BIOMATH calibration protocol [16], HSG-guideline [17] and Water Environment Research Foundation (WERF) calibration protocol [18]. A comparison of strengths and weaknesses of the four systematic calibration protocols STOWA, BIOMATH, HSG and WERF was made in [3]. It was pointed out that the protocols have a lot in common: they all start with a definition of the objective of the calibration, emphasise the importance of collecting data and verifying quality of the data and have similar validation step. However, the protocols differ in the design of the measurement campaigns, the choice of experimental methods and the calibration of parameter subsets. The protocols have mainly been developed for modelling full-scale municipal WWTPs and may therefore not be directly applicable for industrial WWTPs. The STOWA protocol [15] is based on the experience gathered from modelling full-scale wastewater treatment plants in the Netherlands. The aim of the protocol is to be easy to use and minimise the effort on sampling and testing, but provide reliable results. The

7 influent characterisation is based on filtering methods and long-term BOD measurements. Parameter calibration is a stepwise procedure where predefined parameter subsets are manually adjusted in a predefined order. It is emphasised that if major adjustments in the parameter values are necessary, a structural error in the model should be suspected. The BIOMATH protocol [16] has been developed at the BIOMATH research unit in Ghent University, Belgium. The protocol aims to combine the state of the art methodologies for calibration of different processes at the wastewater treatment plants to achieve good prediction capability under variable process conditions. The protocol contains a number of dedicated lab-scale experiments for determining parameter values. It is, however, emphasised that only experiments which provide data for influential parameters, as determined by sensitivity analysis, should be performed. The calibration procedure has two steps: steady-state calibration and dynamic calibration. Parameters having most effect on the long-term behaviour of the model are calibrated in the steadystate calibration with averaged process data. Finally, parameters which are found to be influential on the important model outputs are calibrated with dynamic process data. It is recognized that the calibration is most often performed manually due to the over parameterisation of the ASMs, which yields problems with parameter optimisation algorithms. The WERF protocol [18] was developed based on North American experiences to provide methods and guidance for calibration of the ASMs to municipalities and consulting engineers. The protocol gives excellent advice on specific tasks such as influent characterisation and estimation of nitrification and denitrification parameters. Documentation of the protocol lists several case studies which are helpful for engineers applying the protocol. However, the protocol does not discuss modelling the secondary clarifiers. Moreover, it lacks a clear structure of the complete modelling and calibration procedure. The WERF protocol also suggests the use of sensitivity analysis to determine the most influential parameters. The HSG-guideline [17] has been developed at Austrian, Swiss and German universities with the aim of producing high quality results. HSG-guideline emphasises the importance of carefully documenting the simulation study to allow comparison and reproducibility of results. The guideline does not require using any particular bioprocess model. Therefore it does not give systematic instructions on performing the calibration. However, it does suggest using sensitivity analysis to determine influential parameters. The HSG-guideline differs from the other protocols because it is not a calibration protocol. It is rather a general guideline which when followed should ensure the quality of the simulation study. 2.4.2 Systematic protocols for assessing parameter identifiability and calibration of parameters of the ASMs The activated sludge models are practically unidentifiable for two reasons: the available data is insufficient in quality and quantity, and the model structure does not allow identification of unique values for all parameters [22]. Lack of availability and reliability of sensors for online measurements have been limiting the model identifiability [9,23].

8 Even though the performance and reliability of online sensors have improved substantially, the new sensors have not received widespread acceptance at the existing treatment plants [24]. Even if high quality data were available it wouldn t solve the problem of poor identifiability. The activated sludge models contain non-measurable parameters and state variables. Error in one coefficient may be compensated when determining value for another coefficient and different parameters sets may produce identical results. [9] In most applications, only a small subset of parameters is chosen for calibration due to the poor identifiability of ASM parameters from data [2]. Expert knowledge was and is still used in different steps of the modelling process including the choice of which parameters should be calibrated and how they are calibrated [22]. However, it is possible to apply systematic approaches for selecting identifiable parameter subsets for given model structure and data. Two such approaches can be found in the literature: an approach based on the Fischer Information Matrix (FIM) [19, 21] and an approach based on identifiability measures which combine information on parameter sensitivity and interdependencies [20, 2]. Both approaches use local sensitivity analysis for ranking parameter importance and initial selection of parameter subsets for identifiability analysis. The approach for selecting identifiable subsets of ASM parameters proposed in [19] uses criteria calculated from the Fischer Information Matrices. The model outputs and a priori parameter values are selected first. Then a sensitivity analysis is performed to find a reduced set of parameters which show most sensitivity in the selected outputs. FIM is computed for all subsets of parameters smaller or equal to the size of the reduced set. In practice the exponentially increasing computational effort limits the analysis of large subsets. Subsets of parameters are then ranked by criterion based on the condition number or the determinant of the FIM. The approach proposed in [20] and [2] uses sensitivity analysis for quantifying the influence of individual parameters on selected outputs and collinearity measures for identifiability of parameter subsets. Influential parameters are selected to subsets which are then analysed for parameter interdependencies as quantified by the collinearity measures. The approach aims to find the largest identifiable subset of parameters which can be uniquely identified from the available data. Both approaches are local analyses as the sensitivity functions are calculated at specific parameter values [35]. The sensitivity analysis methodology is presented in detail in Chapter 3.1.4. Even though full search of the parameter space is not made, the local analysis is promising when parameter values leading to acceptable model output are known [20]. This is especially the case with the ASMs, as parameter values for various process configurations are available in the literature. Additionally, the computational cost may prohibit the application of regional methods [20]. As only a subset of all model parameters are adjusted, fixing the parameters a priori which are not included in the identifiable subset can potentially bring bias in the parameter estimates [20].

9 2.4.3 Examples on ASM calibration Some examples from the literature are presented here on calibrating the ASMs for fullscale municipal and industrial wastewater treatment plants Even though calibration approaches based on systems analysis have been developed, most case-studies and practical applications of modelling full-scale WWTPs with the activated sludge models presented in literature apply experience-based and ad-hoc methods for model calibration. Actually, it appears that in the field of activated sludge modelling the only studies using the systematic methods for assessing parameter identifiability and parameter calibration are the studies which demonstrate these methods. In [25], Pardo et al. applied the ASM3 and the STOWA protocol for modelling an activated sludge plant removing organic matter and nitrogen from oil refinery effluents. Influent wastewater was characterised according to the STOWA protocol. Calibration was performed utilising steady-state data and laboratory scale batch tests and by manual parameter adjustment to fit modelled MLSS, effluent COD and nitrogen balance to measured data. No results from dynamic simulations were presented. Methanol is dosed as an external carbon source to denitrification in order to achieve very low nitrate levels in the effluent. The calibrated model was used to study scenarios leading to decreased methanol consumption while maintaining current discharge levels. Hulle et al. [26] applied an extended ASM1 and the BIOMATH calibration protocol for modelling an activated sludge plant treating chemical industry wastewater. Aims of the study were to demonstrate the capability of the BIOMATH protocol in modelling industrial wastewater treatment and use the calibrated model for optimisation of the studied treatment plant. Respirometric batch tests were used in determining central parameters of the model. Additionally, three parameters were manually adjusted to improve the fit of model response to data. The model was validated by running a dynamic simulation with operational data from the WWTP. Finally, the model was used for optimisation of operating strategies of the WWTP and investigation of the impact of production schedules on the WWTP effluent discharges. Vandekerckhove et al. [27] modelled a food industry WWTP with the ASM1. Model parameters were not calibrated, but wastewater characterisation was made according to combination of the STOWA protocol and operator experience. Performance of the model was validated by running a simulation with full-scale treatment plant data. The performance was found to be sufficient even though the model was not calibrated. The model was then used to evaluate options for physical treatment plant upgrade. In [28], a model calibration procedure was proposed, which later became known as the BIOMATH calibration protocol [16]. The procedure was evaluated by calibrating ASM1 for a municipal-industrial WWTP. The calibrated model was to be used in process optimisation focusing on improvement of capacity for nitrogen removal. Laboratory scale experiments were used to estimate values for parameters which were thought to be significant. Local sensitivity analysis was performed for the calibrated model to verify that the calibrated parameters were indeed significant. Tracer tests were performed to

10 characterise hydraulics of the WWTP. Simulation results provided a decent agreement when compared to plant effluent data. Barañao and Hall calibrated ASM3 for an activated sludge plant treating mechanical pulp and paper mill effluents in [29]. The most significant parameters, as determined by a nonspecified method of sensitivity analysis, were calibrated with results from respirometric batch experiments and conventional analyses. The calibrated model was able to fit the measured oxygen uptake rate (OUR) curves. Simulation results with full-scale process data were not presented. Nuhoglu et al. [30] modelled an activated sludge plant treating municipal wastewater with the ASM1. Respirometric batch experiments were used in wastewater characterisation. Four of the ASM1 parameters were manually adjusted from their original values to improve the model fit to process data from the WWTP. The full-scale treatment plant was simulated with the model for a period of 42 days, and the simulation results were in good agreement with the measured values. Satoh et al. [31] applied ASM2 for modelling nutrient removal in pilot scale activated sludge processes treating municipal wastewater. Characterisation of wastewater was chosen rather arbitrarily. Five of the model parameters were manually adjusted to improve the model fit to experimental data. Simulated nutrient and COD profiles in the processes were compared to experimental results, which were in good agreement for the calibrated model. Koch et al. presented calibration and simulation results of ASM3 for Swiss municipal wastewater treatment plants in [32]. Most influential parameters, as determined from local sensitivity functions, were calibrated manually by fitting the model to experimental results of respirometric batch experiments with different substrates under aerobic and anoxic conditions. The same set of parameters was used for simulation of four full-scale and pilot plants. Fit of the simulation results to full-scale plant data was further improved by adjusting some of the parameters. In [33], Wichern et al. reported on their experiences on modelling three full-scale municipal wastewater treatment plants in Germany with ASM3 combined with Bio-P module. Readily degradable COD was determined with respirometric experiments. Parameters calibrated in [32] were used initially. Few of the parameters were manually adjusted for each treatment plant model to increase the correspondence between full-scale simulation results and measured data. In [22], Sin et al. proposed an approach to automate the ASM calibration procedure which is usually carried out manually. The approach was applied for re-calibration of ASM2d for municipal WWTP performing nutrient removal. The model was already calibrated with the experience-based BIOMATH protocol. It should be noted that the available data was of unusually high quality: for example nitrogen measurements were available every five minutes during the 86 day period. The approach begins with selecting the identifiable subset of parameters either by identifiability analysis or expert

11 knowledge. Then a distribution for the parameter values has to be provided. The novel aspect of this approach is the application of Monte Carlo procedure for model parameter estimation. Latin hypercube sampling is used to pick values from the parameter space and a simulation is run with the sampled values. The parameter set which results in the lowest objective function value is chosen. The proposed approach basically does parameter optimisation by covering the parameter space systematically. It would be interesting to see how the Monte Carlo optimisation would compare to evolutionary computing approaches in the same problem. Weijers and Vanrolleghem [19] presented a procedure for systematic selection of identifiable parameters in calibrating the ASMs. The procedure was briefly introduced in Chapter 2.4.2. The procedure was applied on modelling a carousel type wastewater treatment plant with the ASM1. The chosen a priori parameter set consisted of ASM1 default values and values from earlier manual calibration. The most sensitive parameters were first selected by using a local sensitivity analysis. All subsets of two to eight parameters were evaluated with the proposed identifiability criteria to find the best identifiable subset for each size. Identifiability of the subsets was tested with simulated output data and real data. For simulated data with and without noise the parameters were identified accurately. Parameter subsets up to size of five were possible to estimate from real data. However, no simulation results were presented to show the performance of the calibrated model. Finally, the effect of the a priori parameter values was evaluated by sampling the whole parameter space. It was concluded that the a priori selection did not affect the results too strongly. In [2], Brun et al. applied the method of systematic selection and tuning of ASM parameter sets first introduced in [20] for ASM2d. This method was also briefly introduced in Chapter 2.4.2. The modelled process was an experimental lane of a WWTP performing organic carbon, nitrogen and phosphorus removal from municipal wastewater. The process was equipped with four online sensors for measuring phosphate and ammonia in the reactors. The identifiability study started with quantifying uncertainty in the a priori parameter set, which was chosen as the ASM2d default parameters. Also certain parameters were excluded from the study, as they were considered to be better estimated from other experiments. This prior analysis relies heavily on expert judgement. Parameter significance was quantified with a local sensitivity analysis around the a priori parameter values. The identifiability measures, the collinearity index and the determinant measure, were calculated for parameter subsets of different sizes. Subsets up to size of nine parameters were considered identifiable. However, three more parameters were fixed and the final set, which was estimated from the available data, had six parameters. Comparison of simulation results and measured data was presented and it showed good agreement. Finally, the effect of fixing certain parameter values was studied. It was found that the estimated parameter values depend strongly on the values of the fixed parameters, proving that the estimated parameter values were reasonable rather than true or unique. Despite its shortcomings, the identifiability analysis methodology presented in [20] has proven to be a useful tool in analysis of large deterministic models. The methodology has been applied in studies of models other than the ASMs in environmental and chemical engineering [34-37].

12 Machado et al. [21] proposed a procedure for systematic selection and calibration of parameter subsets of the ASMs. The procedure is based on the works of Weijers and Vanrolleghem [19] and Brun et al. [20] but it has some improvements. The procedure was applied on modelling a pilot plant performing nitrogen and phosphorus removal from synthetic wastewater. Local sensitivity analysis is used for initial parameter ranking as in the two previously described methods. Parameter subset selection uses the highest ranking parameters as seeds to which parameters are added one at a time. Identifiability of the subsets is quantified by criteria calculated from the FIM. The procedure was tested on a limited set of data but the results were promising. Improvements over the previously proposed identifiability analyses are the decreased computation time, as it is not necessary to explore all parameter subsets under a certain size, and the decreased need for subjective expert judgement. Ruano et al. [38] evaluated parameter subset selection used in experience-based and systems analysis based approaches with the methods of [20]. It was found that the parameters ranking highest in sensitivity analysis were often omitted from the identified parameter set. Experience-based choices for identifiable subsets were either too optimistic, resulting in poorly identifiable sets, or too pessimistic, resulting in loss of available information. On the other hand, the applied systematic identifiability analysis method suffered from heavy computational demand, rendering it practically useless for analysis of large parameter subsets. 2.4.4 Discussion on the different approaches to ASM calibration Accurate process data with high sampling frequency has to be available for purely process data based ASM identification protocols to make sense. If monitoring of the process conditions is very infrequent, and quality of influent and effluent are measured and recorded only to satisfy the authorities, even small parameter subsets cannot be identified accurately from the data. The quality of ASM calibration would be poor in such a case. For these reasons the experience based calibration protocols described in Chapter 2.4.1 utilise dedicated laboratory experiments to estimate model parameters and to complement the available process data. In successful calibration studies with process data based systems analysis methods the process data has been of exceptionally high quality. The earliest study on systematic selection of identifiable parameter subsets, [19], used effluent and influent data from two day measurement campaign where the data had been sampled every two hours. In [2] online measurements of phosphate and ammonia were available from several compartments of the treatment plant. Three months of process data with ammonium, nitrate and dissolved oxygen (DO) measurements at every five minutes coupled with intensive measurement campaign were utilised in [22] and [38]. Purely process data based ASM parameter identification methods are unlikely to get much success outside academic studies unless online instrumentation is implemented on a larger scale at WWTPs.

13 3. CASE STUDY ON ACTIVATED SLUDGE MODELLING Modelling of biological wastewater treatment has been an increasingly popular topic for research since the publication of ASM1. Activated Sludge Models have been mainly developed for modelling activated sludge plants treating municipal wastewater and therefore may not be directly applicable to industrial WWTPs [10]. However, ASM models with appropriate modifications or extensions have been found useful in modelling and simulation of biological treatment of industrial wastewaters [26, 39-40]. Despite the abundance of literature on activated sludge modelling, pulp and paper wastewater treatment modelling has not received much attention. The objective of the study presented here was to calibrate a modified ASM1 for an activated sludge plant treating pulp mill wastewater and validate the model utilising a long-term simulation of a full-scale WWTP. The primary interest in modelling was to achieve an accurate model of COD and nutrient removal. Applied model is a modified ASM1 introduced by Lindblom [13]. The model is simplified by omitting the biological nitrogen removal by the processes of nitrification and denitrification. Since the wastewater is known to be nutrient deficient, growth limiting effects on heterotrophic bacteria from nitrogen and phosphorus are included in the model. The model was calibrated for the wastewater treatment plant of Stora Enso Fine Paper Oulu pulp mill. Wastewater characterization and model parameter calibration were simple without extensive analytical work. OUR measurements and routinely measured process data were used for these purposes. All modelling and simulation studies were made with data from Stora Enso Fine Paper Oulu pulp mill WWTP. Measurements and wastewater characterisation results from Stora Enso Fine Paper Nymölla pulp and paper mill WWTP are provided for comparison. 3.1 Materials & Methods 3.1.1 Plant & data description All modelling and simulations in this report were made with the wastewater treatment plant of Stora Enso Fine Paper Oulu bleached kraft pulp mill. The wastewater treatment plant is an aerobic activated sludge plant designed for removal of suspended solids and organic carbonaceous material. Wastewater is treated in primary sedimentation before the biological treatment. Primary treatment stages also include ph control, cooling and nitrogen dosing. Nutrients dosing is not actively controlled according to the carbonaceous load to the treatment plant, which results in occasionally too low and too high nutrient to carbon ratios in the wastewater. The activated sludge plant treats an average of 32 000 m 3 wastewater per day. Other technical information and wastewater characteristics can be found in Tables 2 and 3. More detailed description of the treatment plant can be found in [41].

14 Table 2. Technical information of the wastewater treatment plant at Stora Enso Fine Paper, Oulu. Number Area (m 2 ) Volume (m 3 ) Primary clarifiers 1 1963 9100 Equalization basins 1 11 000 Activated sludge basins 1 3400 25 000 Secondary clarifiers 1 2827 12 150 Table 3. Characteristics of the influent, effluent and discharge limits at the wastewater treatment plant of Stora Enso Fine Paper Oulu. Values are averages over the simulation period. Aeration Effluent influent Flow (m 3 /d) 32 000 32 000 COD (mg/l) 1167 541 45 000 BOD 7 (mg/l) 255 14 Tot-N (mg/l) 6.6 2.9 500 Tot-P (mg/l) 1.7 0.7 55 Temperature (ºC) 39 Discharge limits (kg/d) The primary clarifier was not included in this study. Therefore all samples and influent data were collected from the aeration basin influent. Effluent data refers to the secondary clarifier effluent discharged to the sea. Process data from 1 November 2007 to 4 September 2008 was used for the simulations in this study. The influent process data consists of flow rates of influent wastewater, return sludge and wasted sludge, and of total COD, total nitrogen and total phosphorus concentrations. The effluent process data consists of total COD, total nitrogen and phosphorus, and soluble nitrogen and phosphorus concentrations. Flow rates in the data are daily average values. Total COD concentrations are analysed from 24 hour composite samples five times a week for the influent wastewater and daily for the effluent. Nitrogen and phosphorus concentrations are analysed weekly from samples combined from the 24 hour composite samples. Activated sludge and wastewater samples from the WWTP of Stora Enso Nymölla mill were also analysed to characterise the wastewater. At Nymölla mill, activated sludge plant treats on average a total of 81 000 m 3 wastewater per day from sulphite pulp mill and paper mill producing office and graphic papers. The WWTP is an activated sludge plant with two aerated 70 500 m 3 basins and three secondary clarifiers with a volume of 10 000 m 3 each. Influent wastewater is pretreated in primary clarifiers and dosed with additional nitrogen and phosphorus nutrients. Part of the wastewater from bleaching is pretreated in ultrafiltration before activated sludge treatment. 3.1.2 Analytical work Oxygen uptake rate (OUR) measurements were made with sludge and wastewater from the Stora Enso Fine Paper Oulu pulp mill and Nymölla mill wastewater treatment plants. The OUR measurements were made as described in [42] and [43]. Activated sludge was

15 sampled from the aeration basin and wastewater from the aeration basin influent. The measurements in Oulu were made between 8 January and 9 February 2009 once or twice a week. Due to some problems with the OUR equipment useful results weren t obtained until 21.1. With each sample the OUR measurements were made both with the wastewater and potassium acetate as carbon sources. For some measurements a third reactor and oxymeter were used to validate the results of the OUR measurement with wastewater as the carbon source. Specific oxygen uptake rate (SOUR) was calculated by dividing the OUR values with the suspended solids (SS) or volatile suspended solids (VSS) concentration in the reactor. Each sample was analysed also for soluble and total COD, suspended solids and soluble and total phosphorus concentrations. More details on the performed analyses are found in [1]. Dates and analyses made on a given day can be found in Table 4. At the beginning of the measurement campaign there were problems with the equipment and dilution of the sludge. Only measurements which went through without problems are given in Table 4. Table 4. Measurement dates and measurements made on samples from Stora Enso Oulu mill WWTP. ww = wastewater. OUR, ww OUR, ww in another reactor OUR, acetate Total COD, sludge Soluble COD, sludge Total COD, ww Soluble COD, ww SS, sludge 21.1.2009 x x x x x x x x 22.1.2009 x x x x x x x x 26.1.2009 x x x x x x x x x 28.1.2009 x x x x x x x x x 2.2.2009 x x x x x x x x x 4.2.2009 x x x x x x x x x 9.2.2009 x x x x x x x x x Total orto-p, ww Soluble orto-p, ww Soluble orto-p, sludge Total P, ww Soluble P, ww Soluble P, sludge Total N, ww Soluble N, ww SS, ww Ammonium N, ww 21.1.2009 x x x 22.1.2009 x x x x x x 26.1.2009 x x x 28.1.2009 x x x x x x 2.2.2009 x x x 4.2.2009 x x x 9.2.2009 x x x Sludge and wastewater from Nymölla mill were sampled on only two occasions, 19.11.2008 and 4.6.2009. Dates and analyses made on each day can be found in Table 5.

16 Table 5. Measurement dates and measurements made on samples from Stora Enso Nymölla mill WWTP. ww = wastewater. OUR, ww OUR, acetate OUR, OUR, ethanol OUR, 2/3 acetate + 1/3 ethanol OUR, glucose methanol 19.11.2008 x x 4.6.2009 x x x x x x Total COD, Soluble COD, Total COD, Soluble SS, sludge SS, ww sludge sludge ww COD, ww 19.11.2008 x x x x x x 4.6.2009 x x x x x x 3.1.3 Model structure The original Activated Sludge Models were developed with the aim of being able to describe the biological removal of organic carbon and nitrogen from municipal wastewater, and later also biological removal of phosphorus. Municipal wastewater has much higher concentrations of nutrients than is required for the growth of heterotrophic bacteria oxidising organic carbon. Therefore growth limiting effects of nitrogen and phosphorus were not included in the original ASMs. However, in wastewaters from pulp and paper mills the nutrient concentrations are much lower. In many cases it is necessary to add nutrients to achieve complete removal of biodegradable organic carbon. A deterministic model for the biological treatment of nutrient deficient wastewaters should include mechanisms for growth limiting effects of nutrients. An example of such model was presented by Lindblom in [13]. The model presented in [13] has both simplifications and extensions to the original ASM1. Since the model aims to describe only the removal of organic carbon, the model was simplified by removing the biological nitrogen removal processes of nitrification and denitrification and their associated state variables. The model was extended to include description of reduction in sludge production by higher order organisms called protozoa. Even though the presence and significance of protozoa in activated sludge systems is well established [44], this extension should be considered highly experimental as the mechanisms have not been validated. Moreover, OUR measurements and available process data do not have information on dynamics of these processes further deteriorating identifiability of the model. All processes and state variables associated with higher order organisms were omitted from the model applied in this work. As the nitrification process is not included in the model [13] the state variable for nitrate and nitrite nitrogen (S NO ) of the original ASM1 was removed. The active mass nitrogen (X NB ), nitrogen in products arising from biomass decay (X NP ), inert soluble nitrogen (S NI ) and inert particulate nitrogen (X NI ) are explicitly included as state variables in the model unlike in the ASM1. The original ASM1 did not have any state variables for phosphorus. The modified ASM1 has state variables for soluble biodegradable phosphorus (S P ), particulate organically bound phosphorus (X PD ), phosphorus in products arising from biomass decay (X PP ) and two state variables for phosphorus in active biomass (X PB,1 & X PB,2 ). X PB,1 is the minimum phosphorus content in active biomass and X PB,2 is additional

17 variable phosphorus content which is part of the biomass only when excess phosphorus is present. Inert soluble and particulate phosphorus are not explicitly included as state variables. However, they are taken into account in wastewater characterisation. In many cases DO is an important factor limiting the treatment performance in an activated sludge process, because oxygen is required for aerobic degradation of organic material and endogenous respiration. DO is modelled as a state variable in the ASMs and it influences the growth rate of heterotrophic bacteria. Volumetric mass transfer coefficient is used for modelling aeration of the activated sludge basins. DO concentration at the studied treatment plant was in the range of 3-6 mg l -1 throughout the simulated period. DO concentrations in the range of 1.5-2.0 mg l -1 are considered to be sufficient and DO concentrations above 4 mg l -1 do not improve operation of the treatment plant significantly [8]. Therefore in this work DO was not a limiting factor for utilisation of substrate in the simulations. When applying the activated sludge models, the aeration basin is modelled as a continuous stirred-tank reactor (CSTR) or as a series of CSTRs depending on the tank configuration. Component balances for the liquid phase over each CSTR are written for every state variable in the model. Component balances for the liquid phase over each CSTR are written as dc dt i Q = ( Ci, in Ci ) + Ri (1) V where dc i is the concentration of the ith component in mg l -1, Q is the flow rate in m 3 d -1, V is the reactor volume in m 3 and R i is the reaction rate of the ith component in mg (l d) -1. Reaction rates R i in the component balances for each state variable are obtained from the ASM model matrix as R i = n j= 1 y j, i r j (2) where y j,i is the stoichiometric coefficient of the ith component for the jth process, r j is the rate equation for the jth process and n is the number of processes. The stoichiometric coefficients for each state variable and rate equations for each process of the applied model are given in Appendix 1. Complete set of parameter values for the model is given in Appendix 2. The model equations are not described in detail in this report. For complete description of the equations, see the original work [13] or for example [45]. Variable nitrogen uptake Nitrogen is added to the influent of the studied treatment plant to avoid nutrient deficiency in biological removal of organic carbon. However, occasionally the required nitrogen dosage is exceeded or disturbances in the upstream processes cause higher than required concentrations of nitrogen in the influent wastewater. Simulation results in Chapter 3.2.5 show that with no other mechanism for nitrogen removal than the growth