You are on page 1of 8

INTEGRATION OF SEWERAGE, SEWAGE WORKS AND RECEIVING WATER MODELS

J Dudley*, B Tomicic**
*
**

WRc plc, Frankland Road, Swindon, SN5 8YF, UK


DHI, Agern All 5, DK-2970, Denmark

ABSTRACT
Integrated computer modelling of catchments has involved the use of separate programs for the three areas
of sewerage, sewage treatment and receiving water. These three areas have developed their own modelling
approaches, resulting today in the use of different determinands and nomenclature. This has led to confusion
in manipulating data between the three classes of program. The EU-funded TVP programme has led to the
fusion of three of the leading programs in their field into a unified model, with data manipulation being
handled by the programs, relieving users of the need to match data correctly. This programme is being field
tested at many sites within the EU, with the ultimate objectives of improved control of the sewerage/sewage
works system and improved environmental quality in the receiving waters.

KEY WORDS
Sewerage; sewage treatment; receiving water; rivers; modelling; water quality; TVP; STOAT; MOUSE;
MIKE; determinand; integrated; catchment; basin.

INTRODUCTION
WRcs STOAT sewage works modelling package is being integrated with DHIs MOUSE (sewerage) and
MIKE (receiving water) models as part of the Technology Validation project (TVP) Integrated Planning
and Management of Urban Drainage, Wastewater Treatment and Receiving Water Systems, carried out
under the EU INNOVATION Programme. Partners to this project are research institutes, consultants and
end-users from six EU member states:
AMGA Genoa (Italy)
CLABSA Barcelona (Spain)
CUB Bordeaux (France)
Danish Hydraulic Institute (DHI) (Denmark)
Halmstad Municipality (Sweden)
Helsingborg Municipality (Sweden)
North West Water (UK)
Sundsvall Municipality (Sweden)
Urban Hydromatic Centre (Sweden)
Venice Municipality (Italy)
WRc (UK)
This integrated system will facilitate three main levels of usage::
1. The applications run as separate programs, but with a controlling master program that is responsible for
data transfer between the programs using files. This is an extension to the traditional approach to
integrated modelling, where the applications remain independent with no provision for feed-back between
the sewerage, sewage works and receiving water based on model predictions. The advantage over the

traditional approach is that the users do not need to specify the data files to be transferred between the
applications; instead, all they need to do is specify the determinands which are to be transferred from the
upstream to the downstream models.
2. The applications run in parallel as routines controlled by the master program, with data transfer possible
directly between the programs. Because each application can request data being calculated downstream
from its domain it is possible to simulate control schemes based on the sensors located in any part of the
system. As an example, it would now be possible to specify control rules about how in-line sewer storage
will be controlled should the sewage works model predict that sewage works storage capacity has been
reached.
3. The applications run in real time, with additional data being provided on-line from SCADA systems.
Here the models provide a real-time decision support role, where the model predictions are acted upon by
operators. Also the models can become a part of the automatic control system, where the model
predictions are made available as additional inputs to the SCADA system. At this level one of the key
problems is the on-line data assimilation and reconciliation. This primarily concerns the establishment of
correct initial conditions for the on-line simulations, but potentially also the on-line model re-calibration
and measurement data quality assurance. In this process the model variables are compared against data as
it is collected, and the changes are made to state variables, model parameters or the data to reconcile the
data and the model, as best appropriate in the given situation.
In this paper we will outline at the following issues required to provide these levels of integration:

User interfaces
Data compatibility
Data transfer
Data reconciliation

A COMMON USER INTERFACE


The main task of a common user interface for this project is to provide a unified working environment for
the preparation and running of the simulations, as well as the result presentations within the same functional
and graphical milieu. Due to the breadth of the involved programs an ultimate ambition aiming at a unique
interface for setting-up the model throughout the catchment has been set as a long-term goal, outside the
scope of the current project.
The interface for preparation and running simulations consists of a series of dialogues where a full set of
simulation parameters for all involved models (e.g. input files, time steps, simulation periods, etc.) needed
for an integrated simulation can be specified, edited and saved for later use. In addition, the interface opens
direct access to the configuration file, where all relevant information related to the simulation mode and the
data to be exchanged between the involved models is to be specified. An integrated simulation is started and
supervised from inside the common interface.
The result presentation user interface has been designed within the MIKE VIEW - the DHIs results
presentation program designed for sewer and river models. This is a fully user-configurable presentation
program, where results can be presented as animated horizontal or longitudinal profile plots, or as timeseries graphs. The aim is to upgrade the presentation system for presenting the results as the integrated
simulation progresses, both in off-line and on-line modelling applications.

DATA COMPATIBILITY
Historically river modellers have made use of BOD5 as the main determinand for carbonaceous pollution;
while sewage works modelling has widely adopted the use of COD. For sewerage modelling, because of its
close links with river modelling, BOD5 has also been chosen. Integration of the models has therefore
required conversions between BOD5 and COD. But in addition to that, the partitioning of these determinands
has varied between the models. Sewage works modelling has divided COD, by necessity, into biodegradable
and nonbiodegradable COD, and has further divided COD, by necessity, into biodegradable and nonbiodegradable COD, and has further divided COD Into soluble and particulate COD (in some models soluble
is not equivalent to filtered, in other models soluble and filtered are effectively equivalent). The sewerage
models have divided BOD5, if at all, into settleable (30-minute settling test) and nonsettleable, while the
river models have not partitioned BOD5.
One of the strengths of MOUSE TRAP, the water quality model within the MOUSE program, is that it can it
divides component transport into three area:
Advective/dispersive transport
Sediment transport
Sewer reactions
If reactions are not involved then COD can be modelled as easily as BOD, greatly reducing the problems of
data transfer between the different applications. Swedish experience with MOUSE TRAP is that the sewer
systems being studied in the TVP project Integrated Wastewater can be represented without the need for the
sewer reactions. This may not apply for areas where sewage stays for long periods in the sewer, and the
assessment of the appropriate set of sub-models to represent a given sewer will be driven by the match
between the simplest representation and the sewerage data.
For nitrogen, the sewage works models use, as a minimum, total ammoniacal nitrogen, oxidised nitrogen,
soluble organic nitrogen and particulate organic nitrogen. The sewerage models use ammoniacal nitrogen
and ignore the other nitrogen fractions, while the river models also look only at ammonia, but divide
ammonia into ionised and non-ionised forms.
Sewerage and river models commonly ignore phosphorus, while the sewage works models generally look
only at soluble inorganic phosphorus. MOUSE TRAP can support phosphorus, facilitating integration of
these programs.
Finally, suspended solids. Sewage works models divide suspended solids into volatile and non-volatile
fractions, while sewerage models divide them into cohesive and non-cohesive; settleable and non-settleable.
The river models follow the line used by the sewerage models.
Because of this diversity of determinands and the understanding of what these determinands represent the
three models have defined a set of transformations to map different determinands between the programs.
Some examples of the mappings are given below.
MOUSE TRAP to STOAT, COD modelling
f g COD dissolved COD soluble, biodegradable
f (1 - g) COD dissolved COD soluble, nonbiodegradable
(1 - f) g COD dissolved + g COD suspended COD particulate, biodegradable
(1 - f) (1 - g) COD dissolved + (1 - g) COD dissolved COD particulate, nonbiodegaradble
where 0 < f, g < 1

f is the fraction of dissolved COD that is soluble in terms of the biological model, while g is the fraction of
COD that is biodegradable. This simple model has worked well, but the users have the freedom to replace
the transformation equations by their own - should, for example, they prefer to have a constant concentration
of nonbiodegaradble COD, rather than a constant fraction.
STOAT to MIKE 11
g COD soluble, biodegradable + f g COD particulate, biodegradable BOD dissolved
(1 - f) g COD particulate, biodegradable BOD suspended
Biomass Coliforms
where 0 < f, g < 1
f is the ratio of BOD to COD, g is the fraction of particulate COD defined in terms of the biological model
that will behave as dissolved material, and h is a multiplier chosen to use the variability in calculated
biomass concentration as an indicator for coliform counts.

DATA TRANSFER
Data transfer is handled with a master protocol file. This is created by the user interface and contains the
following information:

Programs to be executed. Program sequencing can be in series or in parallel; multiple instances of the
applications can be defined if, for example, the interaction of more than one sewer area or sewage works
with the receiving water is required.

For each urban catchment model there is a list of determinands1 and their associated locations to be
exported between the applications. Each application exports determinands in the native form - for
example, BOD5 or COD. Each determinand-location is assigned a unique identifier.

For each program there is a list of the identifiers known to be exported that will be imported by the
application, and the location where the imported value should be used. This allows controllers in, say,
MOUSE to have available values from STOAT and MIKE. Without this there would be no facility for
integrated control, where the sewer system may have to modify control actions based on the state of the
sewage works.

Transformation routines. These are automatically set up by the interface to ensure that, for example,
BOD5 data calculated in MOUSE is converted to relevant COD fractions before being passed on to
STOAT.

DATA RECONCILIATION
Data reconciliation starts with the pragmatic assumption that the measured data should be given greater
credence than the model predictions. Underlying this is the expectation that the data will have been
preconditioned by the data acquisition software, so that problems such as instrument failure do not result in

Determinands are strictly chemical species being measured. Here we use the term to include other calculated values within the
programs, such as water levels or pump discharge rates.

inappropriate data being passed through to the models. At the time of writing (May 1998) the proposed data
reconciliation method had not been tested with real data this level of integration is scheduled for late 1998.
For the hydraulic aspects of the models data reconciliation is a gradually forcing the model state variables at
the sensor locations towards the measured values. The spatial interpolation between the sensors and the
extrapolation in downstream and backwater directions is automatically achieved by the flow computational
algorithm, which runs in this phase as a model with over-determined boundaries.
The water composition aspects of the modes require a different approach. The measured and predicted
values are compared for acceptability - each measurement is assigned an expected 95-percentile range of
accuracy. If the predictions fall within this range no further action is taken. If the predictions are not deemed
acceptable then the timestep is recalculated with a new value for the expected model parameter that affects
the value. The choice of determinand-response parameter is made from previous off-line modelling. For
example, phosphorus mispredictions in the settled sewage are most affected by changes in the metal salt
concentrations being dosed, while the effluent nitrogen at one site is most affected by the nitrogen uptake in
the biomass, and at another site by the ammonia oxidation rate.

APPLYING THE MODEL


Data collection
Each of the sites has collected data appropriate for the local problem. For all sites this has required flows in
the sewers, but after that water quality parameters gave varied depending on the local consent conditions.
The sewage treatment facilities have varied between the sites, from screening or primary sedimentation
through to activated sludge with tertiary filtration. Not all the sites have been involved with modelling the
receiving water, as for some of the sites the discharge consents from CSO and sewage works are set by law,
and there is no engineering pressure to extend the modelling scope to the environmental impact assessment
of these discharges on the receiving water.
For the sewage works surveys data collection has involved the following:
Two or three day surveys with hourly composite sampling at the inlet and exit to the sewage works
Monitoring total and filtered COD, ammonia, total Kjeldahl nitrogen, suspended solids, total and filtered
phosphorus
Recording internal liquor recycles from sludge thickening or dewatering systems
Recording addition of chemicals for phosphorus removal
Dosing of additional carbon sources for nitrogen removal
Providing annual records of influent and effluent quality. This collection has much missing data,
especially on the influent.
The sewage was characterised using the following assumptions:

Filtered COD soluble COD


Non-filterable COD particulate COD
Biodegradable COD a constant fraction, typically 90%
Volatile suspended solids 75% of the total suspended solids
Soluble and particulate organic nitrogen in a 50:50 ratio

Model construction and calibration


For the sewage works model construction began with site visits and lists of significant treatment processes
and the dimensions. Models of the treatment process were constructed based upon this - an example of one
such is shown in Figure 1. In every instance the processes involved used the Lessard and Beck (1988) model

for primary settlement, either the IAWQ Activated Sludge Model #1 (Henze et al., 1987)or the IAWQ
Activated Sludge Model #2 (Henze et al., 1995) for biological treatment, and the Takacs et al. (1991) model
for activated sludge settlement.
Figure 1 - Example Works Layout

The model was then run with the collected sewage data. The short-term, high resolution data was used to
provide calibration in the following order:

Suspended solids in the effluent


COD in the effluent
Ammonia, where relevant, in the effluent
Phosphorus, where relevant, in the effluent

If data was available for the settled sewage calibration was repeated within an iterative loop to provide
closure on both settled sewage and effluent quality. Sludge production was not recorded and the sludge side
of the model was used as a free parameter to ensure that measured MLSS values and sludge consistencies
from primary tanks were met.
Upon completion of this modelling the calibrated model was run through the yearly data. Because of the
reduced sampling frequency - varying from one daily composite a week to 7 a week - the goodness of fit was
required to follow the trend and the magnitude, without focusing on individual days. An example of the
model predictions is given in Figure 2 and Figure 3 for suspended solids and COD.

Figure 2 - Suspended Solids

AS Lane 1 solids vs STOAT combined solids


200

180

160

140

120

100

80

60

40

20

0
02-Jan-96

21-Feb-96

11-Apr-96

31-May-96

20-Jul-96

08-Sep-96

28-Oct-96

17-Dec-96

05-Feb-97

Figure 3 - COD
200

Effluent COD - Lane 1


180

160

140

120

100

80

60

40

20

0
02-Jan-96

21-Feb-96

11-Apr-96

31-May-96

20-Jul-96

08-Sep-96

28-Oct-96

17-Dec-96

05-Feb-97

CONCLUSIONS
The development of the unified urban catchment modelling system is expected to reduce the work load
required to integrate sewerage, sewage works and receiving water models. This will lead to modelling
scenarios that consider the catchment and permit engineering focus to be on the entire system, rather than on
the three sub-areas. The integration also permits modelling real-time control of the complete system, so that
control objectives in the sewers can include effects caused downstream in the sewage works and receiving
water. The development work will progress towards the inclusion of the integrated models as part of the online decision support and control systems in some cities participating in the project.

REFERENCES
Lessard P. and Beck MB. (1988). Dynamic modelling of primary sedimentation, Journal of Environmental
Engineering 114(4) 753-769
Henze M., Grady CPL., Gujer W., Marais GvR . and Matsuo T. (1987). Activated Sludge Model No. 1,
IAWQ Scientific and Technical Reports No. 1, IAWQ, London
Henze M., Gujer W., Mino T., Matsuo T., Wentzel MC. and Marais GvR. (1995). Activated Sludge Model
No. 2, IAWQ Scientific and Technical Reports No. 3, IAWQ, London
Takacs I., Patry GG. and Nolasco D. (1991). A dynamic model of the clarification-thickening process,
Water Research 25(10) 1263-1271

ACKNOWLEDGEMENTS
The authors thank WRc plc and DHI for permission to publish this paper. The views expressed here are
those of the authors and not necessarily those of WRc plc, DHI, the European Union, or any of the
collaborating cities or companies.