Models for Drinking

Water Treatment
Review State-of-the-Art



Techneau, 06.
May 2006


2006 TECHNEAU
TECHNEAU is an Integrated Project Funded by the European Commission under the Sixth Framework
Programme, Sustainable Development, Global Change and Ecosystems Thematic Priority Area
(contractnumber 018320). All rights reserved. No part of this book may be reproduced, stored in a database
or retrieval system, or published, in any form or in any way, electronically, mechanically, by print,
photoprint, microfilm or any other means without prior written permission from the publisher

Models for Drinking
Water Treatment
Review State-of-the-Art


Techneau, 06.
May 2006
Authors:
Luuk Rietveld (TU-Delft
Jeremy Dudley (WRc)



















This report is: PU




TECHNEAU
TECHNEAU - 1 - May 2006


Contents
Contents 1
1 Introduction 3
2 Description of Modelling Environments 5
2.1 OTTER 5
2.2 Stimela 7
2.3 Metrex 8
2.4 WTP model 9
2.5 TAPWAT 10
2.6 WatPro 10
3 Determinands 13
4 Evaluation 15
References 17
Annex I: Determinands of different models 19


TECHNEAU
TECHNEAU - 2 - May 2006



TECHNEAU
TECHNEAU - 3 - May 2006

1 Introduction
Although drinking water treatment has a long history the mathematical
analysis of these treatment processes is still young. Many flocculation
models are data-driven (Wistrom and Farrel, 1998; Zang and Stanley, 1999;
Baxter, 1999; Yu et al., 2000; Baxter et al., 2001; Baxter et al., 2002) and are
difficult to generalise to other treatment works. Other treatment processes,
such as disinfection and filtration, have been widely studied and the models
are on a sounder basis.
To enhance the ease of use of these models they have been linked together in
flowsheeting programs. Flowsheeting programs were introduced in the
chemical industry in the 1960s, and in wastewater treatment in the 1970s, but
it was not until the 1990s that it was really applied to the water industry. Part
of the reason for this has been the slow growth in the number of available
models, so that there was no need to have a program that would allow the
different models to be readily linked together. Another reason has been that
mathematical modelling has largely been seen as an academic activity, with
real water treatment plants designed using rule-of-thumb approaches by
engineers with years of experience. With the emphasis on water quality,
rather than economic efficiency, this approach produced working, if
somewhat conservative, plants.
With growing pressures on water treatment there is now a greater need to
optimise water works, whether to increase throughput, reduce operational
costs, or minimise capital expenditure. One tool that can assist with this is
water-treatment flowsheeting packages. Although programs for individual
treatment processes, of various quality and in different programming
languages, exist there are only a few packages for the simulation of the entire
treatment plant. This report focuses on the State-of-the-Art of these modelling
platforms for drinking water treatment. The programs that are considered are
known from the literature, found through Internet searches or obtained by
informal contacts.



TECHNEAU
TECHNEAU - 4 - May 2006


TECHNEAU
TECHNEAU - 5 - May 2006

2 Description of Modelling Environments
The following drinking water treatment modelling platforms are selected for
description and evaluation:
OTTER, from WRc, a project partner and for which full source code is
available
Stimela, from TU Delft, a project partner and again for which full
source code is available
METREX, from Duisburg University. Although not a project partner
they are collaborating through TU Delft and have agreed to make
available their source code.
The following programs are also discussed, although source code for these
programs is not available:
WatPro
WTP
TAPWAT

2.1 OTTER
OTTER is a PC based modelling package designed to dynamically simulate
the performance of water treatment works (Head et al., 2002). Development
of OTTER as combined package began around 1996, with the earliest part of
OTTER (the carbonate chemistry module) dating back to the early- to mid-
1980s. OTTER development has continued since, with the last public release
of OTTER in 2003. OTTER has seen use throughout the world, but
predominantly in the UK and USA.
The program can be used to simulate individual treatment processes or the
whole treatment plant. It enables process scientists and plant operators to
optimise the response of the works to changes in the raw water quality, plant
throughput or process operating conditions. Typical uses of the software
include operational decision support, works optimisation, plant design and
operator training. Version 2 of OTTER includes:
Chemical floc formation and pH adjustment
Clarification (floc blanket clarifiers, dissolved air flotation,
sedimentation tanks, lamella settlers)
Rapid gravity filtration
Granular activated carbon adsorption
Ozonation
Disinfection
Sludge treatment
OTTER models the formation and removal of a wide range of water quality
parameters, including general parameters such as turbidity and colour,
organic parameters such as DOC, inorganics such as bromate, pesticides and
microbiological indices. A full list can be found in Annex I.
OTTER supports user-added models using a dynamic link library. Thus many
programming languages can be used to develop new OTTER models,
additional to those provided in the basic suite, reducing the effects of vendor

TECHNEAU
TECHNEAU - 6 - May 2006

lock-in compared to simulation packages that use proprietary languages.
However, the experience of WRc has been that few users have the skill,
experience or interest to develop models of their own, preferring instead that
such models as they need be developed for them and included in the core
OTTER system.
OTTER itself is written in a combination of Microsofts Visual Basic, for the
user interface, and Fortran code, for the numerical models. The Fortran code
has demonstrated its adaptability, as different Fortran compilers have been
used during OTTERs history. The Visual Basic code is, of course, tightly
coupled to Microsoft and only available through Microsoft. However, for the
end user there is no need to pay for copies of Visual Basic or Fortran. An
example of the visual interface is shown in Figure 2.1.


Figure 2.1 OTTER flowsheet

The solution approach is modular sequential. That is, at each timestep, all the
differential equations in a process unit are integrated across the range T to T +
T. Then the next unit downstream is solved, being passed as its water flow
and quality the value for the upstream process at T, treated as piece-wise
constant across the integration interval. Because of this modular sequential
solution method a loop finder, with reordering of the calculation order to
minimise the effects of loops, and iteration around the loops, is automatically
provided. The calculation order can be altered by the user.
Several studies have been done using OTTER at waterworks (Butler, 1998;
Gallis, 1999; Giraudet, 2002a, b; . Guo and Sankararamkrishnan, 2003).
Generally these have all been successful, but have highlighted the relatively
large data requirements for successful calibration.



TECHNEAU
TECHNEAU - 7 - May 2006

2.2 Stimela
Stimela is an environment where different drinking water treatment
processes can dynamically be modelled. The Stimela models are developed in
Matlab/Simulink. Matlab is a commercial modelling environment available
from The MathWorks. It comprises a language geared around matrix
methods and is extensively used in academia and as a prototyping
environment. There are many add-ons to Matlab to simplify its use in a wide
number of areas The MathWorks claim that there are several hundred
commercial toolboxes available. The most widely used toolbox is the
simulation toolbox, Simulink. Matlab/Simulink has been used by both
Metrex and Stimela, and has also been used by a wastewater modelling
environment, Simba, which was the inspiration for Stimela and Metrex
electing to use Matlab/Simulink.
Partial differential equations are converted into ordinary differential
equations, usually by using either finite differences or a series of mixed tanks.
The resulting ordinary differential equations are numerically integrated so
that variations in time and space can be followed. The models of individual
processes are situated in a model library and can be connected to each other,
forming a complete treatment train (see Figure 2.2). In this way, the effect of
operational changes in preceding treatment processes can be evaluated.
Because Matlab/Simulink is used, the models are easily accessible, the
structure is open and flexible and all routines, toolboxes and visualisation
techniques of Matlab/Simulink can be used (Helm and Rietveld, 2002).
However, this does require that end-users purchase a copy of
Matlab/Simulink, and for non-academic users the purchase cost is high. In
addition, the use of Matlab/Simulink results in a user interface that is more
difficult to learn than more customised simulation software.
Stimela includes the following processes:
Aeration (cascades, towers, plates, spayers)
Filtration (single layer, double layer, continuous, biological)
Granular activated carbon filtration
Softening and conditioning
Ozonation (bubble column and contact chambers)
Stimela models dissolved compounds such as gasses (CH
4
, CO
2
, O
2
, O
3
),
inorganic compounds (HCO
3
-
,NH
4
+
, CO
3
2-
, Ca
2+
) and organic compounds
(DOC, organic micropollutants, UV
254
, AOC). In addition floc removal by
filtration is modelled. A full list is available in Annex I.
A Stimela model consists of a raw water block, followed by one or more
process blocks, control blocks and graphical output blocks. The raw water can
be specified as constant values, or a datafile can be specified to provide time-
varying values. The connecting line between blocks transfers the values of the
water quality parameters at every time step. By opening the process block,
design, and calibration parameters can be inserted. The design parameters
(e.g. tank volume) are fixed. Calibration parameters are water-specific and
must be obtained through experiments. Calibration is executed with obtained
measurement data of a pilot or full-scale plant. Calibration parameters can be
floc density, filtration coefficients, Freundlich- constants, gas transfer
coefficient, etc. Control blocks can be used to further manipulate the process
blocks, varying calibration or operational parameters.

TECHNEAU
TECHNEAU - 8 - May 2006



Figure 2.2 Example of a treatment train in the Stimela environment

After specifying all parameters the flowsheet model can be run, choosing the
integration method, the step size and the simulation time. After simulation,
graphical output is obtained by opening the output block. The calculated
values can be compared with measured data. Typically, the output consists of
water quality parameters that are relevant for the process and other data that
describe the state of the process, such as filter head loss, degree of saturation
of activated carbon, grain size of pellets in a softening reactor, etc.
Stimela is available on the web, at www.stimela.com, where it may be freely
accessed.

2.3 Metrex
Metrex was developed at the University of Duisburg, Germany (Mlzer and
Nahrstedt, 2002). The simulation tools combine analytical and numerical
models of common treatment steps used in surface water treatment:
Microstraining
Ozonation
Floc formation
Sedimentation
Rapid filtration,
Granular activated carbon filtration
Biodegradation
Disinfection
The emphasis is on particle removal (particle size distributions are
considered) and ozonation (oxidation of dissolved organic carbon, iron and
manganese, formation of bromate). Simulated plants can be configured in any
combination of treatment steps. A graphical user interface assists with setting
up the underlying mathematical models and required data.

TECHNEAU
TECHNEAU - 9 - May 2006

Two levels of simulation exist. The first is designed to simulate the whole
treatment process in operation mode, while the second provides support in
designing and dimensioning single process steps. The software concept is
based on an object-oriented model using Java technology, World-Wide-Web
(WWW) browsers and the simulation software tool Matlab/ Simulink. Each
process step is designed with a uniform structure. Water compounds are
characterized by their concentrations and in some cases additionally by their
treatability (e.g. biodegradability, adsorbability on activated carbon). Reports
of the simulation results, including all water parameters, are generated
automatically in Hyper-Text-Mark-up-Language (HTML) format.
Documentation is also given using HTML files. Mechanistic models are used.
In many cases qualitative knowledge exists about the range of values of the
parameters and their dependencies on other known variables. This enables
the user to estimate values for parameters using fuzzy-linguistic modelling.

2.4 WTP model
The Water Treatment Plant (WTP) model was originally developed by the
Environmental Protection Agency, in support of the Disinfectant/Disinfection
By-products (D/DBP) Rule (Harrington et al., 1992). The models describe the
processes:
Coagulation/flocculation
Settling
Filtration
Granular activated carbon filtration
Softening
Membrane filtration
Chlorination
It is based on empirical relations obtained from regression analysis. It was
prepared with the understanding that the predictions should reflect typical
average performance values, and is focussed on the removal of natural
organic matter (NOM), the formation of DBPs and disinfection. It is not to be
construed that the results from the model will necessarily be applicable to
individual raw water quality and treatment effects at specific municipalities.
This model does not replace sound engineering judgement based on site-
specific treatability data to evaluate the best manner in which to address the
requirements of the Surface Water Treatment Rule (SWTR) or potential
D/DBP Rule. It is understood that one limitation of the model is the extent of
the database availability to verify model predictions. In a desire to
systematically improve the overall predictive capability, the intent of the
model is to solicit public comment on the usefulness and relative accuracy of
the predictions on a case-by-case basis. The WTP model includes a method to
enter laboratory analysis so that a comparison can be made to the model
predictions. In 2000 the WTP model was modified. Existing algorithms were
updated and new process algorithms were added, especially related to
inactivation of microorganisms, formation of DBPs and the decay of
disinfectants.

TECHNEAU
TECHNEAU - 10 - May 2006

2.5 TAPWAT
TAPWAT (Tool for the Analysis of the Production of drinking WATer) wass
developed by the Netherlands Institute of Public Health and the Environment
(RIVM), with the following goals:
To predict on a global scale, the quality of drinking water (including
health risk levels by micro-organisms) given a certain raw water
quality
To determine the probability of occurrence of pathogenic
microorganisms and DBPs in the product of a treatment plant
To advise the drinking water inspectorate by reviewing new or
renewed production plants especially concerning public health risks
The program supports percentage-removal based, empirical and mechanistic
models for all process units. The choice as to which level of complexity to use
is then made based on available data and user confidence with mechanistic
models outside their calibration region. The program was developed
according to RIVM standards for information infrastructure and modelling
(Versteegh et al., 2001).
Percentage removal models were developed using information from the open
literature, reports, data from production plants and expert judgement to
provide design ranges of percentages for removal of microorganisms and
compounds by the selected treatment processes.
Process models for the formation of DBPs by chlorination and ozonation were
tested using data from full-scale plants in the Netherlands.
The beta-distribution is used for statistical analysis of data of production
plants, especially for the removal of TOC and turbidity by floc formation, floc
removal and activated carbon filtration. The distribution of the product water
quality parameters is determined by the distribution of the influent water
quality parameters and the distribution of the removal percentages. For this,
Monte Carlo techniques are used.

2.6 WatPro
WatPro is supplied by Hydromantis Inc . It is a steady-state water treatment
modelling program, with a focus on disinfection and disinfection by-
products. Although other aspects of water treatment processes are supported
these are of lesser significance within the packages scope. The information in
this section is taken from the WatPro user guide (Hydromantis, 2004).
Supported treatment processes are:
Measurement point: This is used to adjust the water quality at any
point to measured values. It is specifically intended to allow raw
water quality to be adjusted to reflect any unmodelled treatment
processes, or correct for any deficiencies in the model output
compared to experience
Flocculation
Settling basin this process does not appear to model settlement;
rather, allowing the user to specify the outlet turbidity
Filtration simple models, where the user specifies the percentage
removal of TOC and UV254, and optionally the effluent turbidity

TECHNEAU
TECHNEAU - 11 - May 2006

GAC this appears to be an incomplete model, as it uses a Freundlich
isotherm to describe removal of TOC and UV254 but requires that the
user provide only one of the two parameters in the Freundlich model,
the exponent n
Membrane again a simple model, where the user specifies the
percentage of water produced through the membrane (the remainder
(the waste stream) is treated as the concentrate and optionally the
effluent turbidity
Contact tank for chlorine disinfection
Ozonation ozone disinfection
WatPros strength is in the prediction of chlorination by-products, using
published US EPA correlations. It is otherwise a poor example of the state of
the art in water treatment, modelling all other water treatment processes by
the user either specifying a percentage removal, which is independent of
water quality or flow, or an outlet turbidity from the process, again
independent of water quality or flow. Because of these serious deficiencies the
use of WatPro for any purpose other than modelling disinfection should be
avoided by inexperienced users.




TECHNEAU
TECHNEAU - 12 - May 2006


TECHNEAU
TECHNEAU - 13 - May 2006

3 Determinands
The appropriate selection of determinands (water quality parameters) in
water treatment should ideally meet the following criteria:
Be based around what is likely to be measured, and not over-whelm
the user with a large number of unnecessary determinands
Have the ability to add new determinands, as water quality
requirements change for example, it is likely that modelling the fate
of endocrine-disrupting chemicals may be of greater interest in future
Provide a comprehensive list of determinands, so that the user does
not have to locate any needed physical or chemical property data
Clearly these requirements impose some conflicts. For OTTER, Stimela and
METREX an overview of the determinands is given in Annex I.
The OTTER approach has been to provide a large, but organised, list of all
determinands that can be modelled. The addition of any new determinands
would potentially require modifications to all unit processes. The OTTER
determinands dialogue screen is shown in Figure 3.1.


Figure 3.1 OTTER determinands

Stimela, on the other hand, provides a smaller list of determinands, and
allows the user to make active only those parameters that are actually
required. However, there is less emphasis on trying to categorise the
determinands. The Stimela determinands dialogue menu is shown in Figure
3.2.


TECHNEAU
TECHNEAU - 14 - May 2006


Figure 3.2 Stimela determinand dialogue

Metrex also categorises the determinands. Its initial determinand dialogue is
shown in Figure 3.3


Figure 3.3 Metrex determinand dialogue



TECHNEAU
TECHNEAU - 15 - May 2006

4 Evaluation
Each of the water treatment plant models discussed above has its specific
characteristics. OTTER contains models for most commonly encountered
processes and less conventional processes may require development of a
suitable mathematical model (Butt and Head, 2002). OTTER is a well-
developed model for design purposes in particular. It is graphically good and
it is easy to build up a treatment plant. Visualisation and calculations are
separated. It is well documented and models of many different conventional
treatment steps are available. The main weak point is that it is a compiled
program and therefore it is difficult for general users to add new process
models.
Stimela uses Matlab for its calculations and can also use its functionalities
(integrator, loop detector, debugging). Stimela has a Web interface, but it is
limited to single processes. The main purpose of Stimela is to support
research and development and control applications. Therefore, it focuses on
model development, programming is open and structured and graphical
output is flexible. Although documentation is limited several scientific
publications (journals, congress proceedings, PhD thesis) are available
describing the results of the modelling efforts. Operating within
Matlab/Simulink, many of the models are interpreted and simulating large
systems the simulations can require extensive calculation time. It is also
difficult for general users to add new process models without an extensive
knowledge of Matlab/Simulink and the Stimela framework, plus, of course,
licenced copies of Matlab and Simulink.
Metrex is being developed mainly for particle removal, but has as an
advantage that it is written in an accessible language and exchange and
development is easy. Metrex, like Stimela, makes use of Matlab. Simulation
on different levels is possible and many different water quality parameters
(determinands) can be incorporated, but not all in use. However, Metrex is
currently moribund.
WTP model was developed to simulate the general case rather than the site-
specific case. Some drinking water treatment plant operators may be tempted
to use this model as a substitute for site-specific studies. However, the output
from the model is not intended to, nor should it, replace sound engineering
judgement based on bench-, pilot-, and field-scale treatability studies for
specific waters (Harrington et al., 1992). In fact, the model is mainly used for
evaluation of design rather than operational optimisation studies.
With TAPWAT stochastic and simple process modelling can be performed for
pathogenic microorganism removal and disinfection by-product formation.
The model is not intensively tested and still in a development phase. It is not
used outside RIVM.
WatPro is effectively a generalisation of WTP, and its algorithms replicate
those found in WTP and the disinfection modules of OTTER and Stimela.


TECHNEAU
TECHNEAU - 16 - May 2006




TECHNEAU
TECHNEAU - 17 - May 2006

References
Baxter, C.W., Shariff, R., Stanley, S.J., Smith, D.W., Zhang, Q., and Saumer,
E.D., 2002, Model-based advanced process control of coagulation, Water
Science and Technology Vol. 45(, No. 4-5), pp. 9-17, 2002.

Baxter, C.W., Stanley, S.J. and Zhang, Q., 1999, Development of a full-scale
neural network model for the removal of natural organic matter by enhanced
coagulation, J. Water SRT-Aqua, Vol. 48(4), No. 4, pp. 129-136, 1999.

Baxter, C.W., Zhang, Q., Stanley, S.J., Shariff, R., Tupas, R-R.T and Shark,
H.L., 2001, Drinking water quality and treatment: the use of artificial neural
networks, Can. J. Civ. Eng. 28 (Suppl. 1): 26-35, 2001.

Butler, M, 1998, The analysis and modelling of two water treatment works,
MSc thesis, Birmingham University, Water Resources Technology and
Management.

Butt, G. and Head, R., 2002, Development of a prototype treatment works
management based on real-time process simulation, Water Science and
Technology: Water Supply Vol. 2(, No. 1), pp 81-86, 2002.

Gallis, Z, 1999, Modelling and simulation of water treatment plants, MSc
thesis, Imperial College, Department of Civil and Environmental Engineering,
University of London.

Giraudet, T, 2002a, Development of OTTER software for the Iver Water
Treatment Works, Vivendi Water internal report.

Giraudet, T, 2002b, Development of OTTER software for the Walton-on-
Thames Water Treatment Works. Vivendi Water internal report.

Guo, QGG and N Sankararamakrishnan, 203, Development of a numerical
model for assessing the impact of raw water quality on conventional drinking
water treatment, New Jersey Department of Environmental Protection,
Trenton, New Jersey.

Harrington, G.W., Chowdhury, Z.K. and Owen, D.M., 1992, Developing a
computer model to simulate DBP formation during water treatment, Journal
AWWA, Vol. 84( No.11), pp. 78-87, November 1992.
Head, R., Shepherd, D., Butt, G. and Buck, G., 2002, OTTER mathematical
process simulation of potable water treatment, Water Science and
Technology: Water Supply Vol2(, No 1), pp 95-101, 2002.

Helm, A.W.C. van der and Rietveld, L.C., 2002, Modelling of drinking water
treatment processes within the Stimela Environment, Water Science and
Technology: Water Supply Vol 2(, No 1), pp 87-93, 2002.


TECHNEAU
TECHNEAU - 18 - May 2006

Hydromantis, 2004, WatPro Version 2 Users Manual, supplied with
demonstration version of WatPro.

Mlzer, H-J and Nahrstedt, A., 2002, Modellierung mehrstufiger
Trinkwasseraufbereitungsanlagen mittels eines expertensystem-basierten
Simulationsmodells (Metrex) am Beispiel von oberflchenwasser., IWW, 2002.

Versteegh, J.F.M., Gaalen, F.W. van, Rietveld, L.C., Evers, E.G., Aldenberg,
T.A. and Cleij, P., 2001, TAPWAT: Definition, structure and applications for
modelling drinking-water treatment, RIVM report 734301019, 2001.

Wistrom, A. and Farrel, J, 1998., Simulation and system identification of
dynamic models for flocculation control, Water Science and Technology,
Vol. 37(, No. 12), pp 181-192, 1998.

Yu, R.-F., Kang, S.-F., Liaw, S.-L. and Chen, M.-C, 2000,., Application of
artificial neural network to control the coagulant dosing in water treatment
plant, Water Science and Technology, Vol. 42(, Nos. 3-4), pp 403-408, 2000.

Zhang, Q. and Stanley, S.J., 1999, Real-Time water treatment process control
with artificial neural networks, Journal of Environmental Engineering,
February 1999.

TECHNEAU
TECHNEAU - 19 - May 2006

Annex I: Determinands of different
programs
Otter Stimela Metrex
General:
pH, Temperature, Turbidity,
colour, Total hardness,
Alkalinity, EC, TSS, SSS, FSS,
Cl, ClO3, ClO4

pH, Temperature, Turbidity,
Total hardness, Alkalinity,
EC, TSS

Ph, Ks, Kb, alkalinity
Metals:
Al, Mn, Fe

Fe

Al, Mn, Fe
D.O. (mg/l) Gasses:
CO2, O2, O3, CH4

O2, O3, N2
Inorganics:
Ammonia (mg/l), Nitrate
(mg/l), Nitrite (mg/l),
Chloride (mg/l) Chlorite
(mg/l), Chlorate (mg/l),
Bromide (g/l), Bromate
(g/l), Sulphate (mg/l),
Orthophosphate (mgP/l)

NH4, NO2, NO3, PO4

BrO3, Br, Ca, Cl, F, K, Mg,
NH4, NO2, NO3, OH, PO4,
SO4, HCO3, CO3
Organics:
TOC (mg/l), DOC (mg/l),
POC (mg/l), THMs (g/l),
THMFP (g/l), HAAs (g/l),
AOC (g/l)

DOC, TOC, UV254, AOC

DOC, UV254, UV436
Pesticides:
Simazine (g/l)
Chlortoluron (g/l), Diuron
(g/l), Isoproturon (g/l),
MCPA (g/l), MCPB (g/l),
Mecoprop (g/l), Diazinon
(g/l), Chlorfenvinphos
(g/l), Propetamphos (g/l)
flexible flexible
Microbiological:
Coliforms (No./ml), E.Coli
(No./ml), Viruses (No./ml),
Heterotrophs (No./ml)

Others:
Chlorophyll-a (g/l), Taste
(Number), Odour (Number),
Particle Size (m), Particle
Count (No./ml)

Particles