You are on page 1of 17
STOAT - DEVELOPMENT AND APPLICATION OF A FULLY DYNAMIC SEWAGE TREATMENT WORKS MODEL Dr Jeremy W. O. Dudley, Deborah A. Bryan, and Dr Brian Chambers WRe ple, Frankland Road, Blagrove, Swindon, Wiltshire, SNS 8YF, UK ABSTRACT ‘The 1980's saw the increasing realisation that it was both feasible and desirable to design and operate sewage treatment plants and sewerage systems in an integrated manner. In the UK, the Urban Pollution Management (UPM) programme focused these thoughts and resulted in the specification and development of a fully dynamic sewage treatment works simulation model, now known as STOAT. This paper describes the current status of STOAT and examines its use as part of an integrated design package for urban wastewater systems. The latter part of the paper goes on to consider the potential for operational use of the model in real time control. STOAT has been developed in modular form with process and transport modules based on a combination of well tried and tested simulation models (eg. the WRe Activated Sludge model) and newly developed units. At present the model can represent most commonly occurring treatment facilities incorporating storm tanks, primary sedimentation, secondary treatment by both activated sludge and ‘biological filters, oxidation ditches, final clarifiers and tertiary filtration and both aerobic and anaerobic sludge digestion. The software package consists of a flexible menu and picture driven interface operating through Microsoft Windows. It is fully compatible with the other UPM tools (MOSQITO for the sewer system; MIKE 11 for the river) to provide an overall integrated modelling package. These models have been used together on several case studies, One of these is reported here, where STOAT is used 10 demonstrate the effect of addition of storage in the sewer system on the sewage treatment works and on the input to the receiving watercourse from the works and combined sewer overflow. ‘The paper concludes by looking at the potential application of STOAT for real time operational control, illustrating how the model can assist both in optimising operation of the treatment plant and to manage the sewer system to complement the weatment plant capabilities. KEYWORDS Dynamic models, sewage treatment, process modelling, real-time control, environmental impact, urban pollution management. INTRODUCTION In 1987 WRc began the development of STOAT - Sewage Treatment Operation and Analysis over Time. STOAT was seen as part of a triad of models around which the UK’s Urban Pollution Management (UPM) programme would be built. The triad would comprise a sewerage quality model - MOSQITO; a sewage works process model - STOAT; and a river quality model - MIKE 11. These were not seen as definitive tools, but as models with which to acquire experience. The aim was to examine integrated environmental modelling with the ability to study the interplay and effects of decisions made on the design and operation of the sewerage system and sewage works on the receiving waters. DEVELOPMENT This phase of the UPM programme was completed in 1992. STOAT existed as a research-based program and had been used to prove the integration of the modelling triad, with data transfer between the three models. However STOAT was difficult to use and WRc began a programme of further development. The first phase of development was to provide a better user interface, built around the popular Windows system. Following this the process models were improved, after which the interface was further developed. STOAT will continue to be developed, new processes are added as the modelling knowledge becomes available and the interface is enhanced to meet users’ requirements. PROCESS MODELS STOAT at present has the following processes: * Storm tanks . Primary tanks * Wet wells * Activated sludge - including oxidation ditches, N and P removal systems * Activated sludge settling tanks * Trickling filters - nitrifying and carbonaceous : ‘Humus tanks * Anaerobic sludge digestion * Thermophilic aerobic sludge digestion In addition there are both process control and simple models to allow the user to specify empirical equations for treatment processes. The main processes listed above support the dynamic modelling of sewage treatment works. ‘These models work with the determinands commonly used by the sewerage and river quality models - BOD, suspended solids, and ammoniacal nitrogen. In addition nitrate, dissolved oxygen and organic nitrogen can be included in the models. For biological phosphate removal the models also require soluble phosphate and volatile fatty acids as determinands. The models were mainly validated in the STOAT and UPM development. The activated sludge model was developed by WRe in 1972 and has been extensively used since the early 1980s. The biological phosphate removal model has been added recently and is based on work leading up to the IAWQ Activated Sludge Model #2, modified for BOD. The P-removal model is being developed in collaboration with Imperial College, London, and will be published as a PhD thesis in 1995, All WRe biological models incorporate the concept of viability. That is, the biomass is divided into two parts. The first part is viable and can grow while removing BOD or ammonia. Some of these viable bacteria decay to nonviable bacteria. Unlike the [AWQ models the WRe models assume nonviable bacteria still possess biological activity as they contain active enzymes, although they cannot grow. Nonviable bacteria decay further into dead cells and inert solids, when there is no longer any biological activity. The storm and primary tank models were developed by Bruce Beck and his team of workers at Imperial College, London. These models were validated at Norwich sewage treatment works in the early 1980s and have since been further validated by Bruce Beck’s (Lessard and Beck, 1989, 1988) team at other sewage works and by WRe within the UPM programme. The thermophilic aerobic sludge digestion model was developed by WRe and validated on several UK digesters. The anaerobic digestion model was also developed by WRe. Several papers describing this model have been published by Mosey (Mosey, 1983). Because of the complexity of Mosey’s model STOAT also incorporates’ a simple first-order decay model which can be readily calibrated from a few laboratory tests. The trickling filter model is another model developed by WRe and validated during the UPM programme. Like the activated sludge model it incorporates a distinction between viable and nonviable biomass. This model has been used 10 model BOD-only removal; a tertiary nitrifying filter; and the switch over from BOD removal to nitrification in a deep-bed filter. USER INTERFACE ‘The user interface was developed in the Excel spreadsheet. This allowed rapid development of the interface followed by evaluation and suggestions for improvements. In January 1994 STOAT was released in this version. Meanwhile development of the interface continued. The interface has now been written using Visual Basic and this enhanced version of STOAT is scheduled for release in August 1994, Users have access to a screen on which they can design their own sewage works. Processes are selected from a graphical toalbox and placed where required on the screen. These processes are then connected in the required order, with the program carrying out checks to ensure that there are no logical inconsistencies in the connections (for example, the output from one process cannot be connected directly to the output of another process). Having constructed the works layout the sizes of tanks, flowrates, etc., must be specified. The user selects any process and is given a menu of data. This allows a choice of models - for example, should the activated sludge model be the standard or P-removal versions - as well as data fields. Where relevant default data are available it is given to the user, who can then choose to alter the values Gs necessary. When all data have been entered the process can now be ar nulated. Sewage flow and quality data can be taken from MOSQITO and SXported back 10 MOSQITO. MOSQITO contains conversion programs for SOSQITO to MIKE 11. STOAT will soon also include conversion modules for the MOUSE software family. The mode! runs interactively. Users can select which variables they would like to see displayed on the screen while the model runs, and can further select ‘Rodel parameters to be altered during the simulation, This approach allows the rapid evaluation of a variety of "what if" scenarios. The rapid feedback allows the users to stop simulations that become infeasible, while the. interactive Control allows the parameters to be changed if it is required to simulate, for example, sudden process failures. ‘The results from the simulation are stored in a database for_subsequent manipulation, Post-processing can produce graphs for any of the flowstreams, Uhether within the works or the final effluent, as well as summary statistics for tach process and flowstream, The database facility allows users fo extract data for fanher calculations. STOAT also allows data to be sent to the Excel spreadsheet so that spreadsheet analyses can be carried out. USING STOAT FOR UPM For the scenario given here STOAT was linked to MOSQITO and the Stochastic Rainfall Generator to examine the effects of adding storage within a sewerage system on the sewage treatment works and receiving watercourse. The Rainfall Event The storm used begins at 09:00 hrs when the highest foul flows are in the sewerage system, 10 mimic a worst case scenario. The rainfall event used was generaied by the Stochastic Rainfall Generator (SRG). This model can generate a representative hourly rainfall series for any given time and specified location inthe UK aaa forms part of WRe's STORMPAC program. The synthesised series can be used as a substitute for a historical local hourly rainfall récord. ‘The event chosen for this demonstration has the following characteristics: Approximate return period (based on spill volume) = 9 months, Storm duration = 6 hours Total rainfall = 26.3 mm Maximum intensity = 6.8 mm/hour Urban Catchment Weiness Index value = 144 MOSQITO Modelling of the Sewerage Catchment ‘An existing MOSQITO model, of the Litlleborough sewerage system, was used to provide input to the STOAT model. There is one combined sewer overflow (CSO) in the system. The MOSQITO model was mun in 3 configurations, using a box culvert with 0, 1500 and 3000 m’ of storage respectively, downstream of the CSO. Downstream of the box culvert a dummy tank was modelled, which did not spill, but restricted pass-forward flow to the Formula A limit. Formula A is defined as: Formula A = DWF + 1.36P + 2E m/day where: DWF =PG+I+E P = Catchment Population G = Water consumption per head m°/day I= Infiltration m’/day E = Trade effluent m'/day Table 1 shows the characteristics of the modelled catchment. TABLE 1 CATCHMENT CHARACTERISTICS ee Population 17500 Combined sewered area (ha) 220 Impermeable area (ha) 93 Number of CSOs 1 ‘Average dry weather flow (V/s) B ‘Average dry weather BOD (mg/l) 194 ‘Average dry weather SS (mg/l) 400 ‘Average dry weather NH,-N (mg/l) 12 MOSQITO was run, using the rainfall event as input, for the 3 culvert sizes. The model was also run with dry weather inputs only. The outfall flow and quality from the four runs were used as input to STOAT. The volume of flow and BOD and ammoniacal nitrogen loads spilling from the CSO to the river were recorded. Table 2 shows these results. TABLE 2 MOSQITO RESULTS FOR CSO PERFORMANCE Tank storage (mr) 0 1500 3000 Spill volume (m°) 11019 9908 8495 BOD spill load (kg) 544 367 290 Maximum BOD concentration (mg/l) 183 93 71 NH,-N spill load (kg) 93 15 61 Maximum NH,-N concentration (mg/l) 22 19 19 It can be seen that there is a considerable reduction in BOD spill load and maximum concentration when 1500 m° storage is added to the system. This is” mainly due to the containment, of the first foul flush within the system. Increasing the storage to 3000 m® gives only a slight reduction in the pollution load. Clearly from an operators viewpoint there is an optimum storage size for reduction of the CSO spill load, beyond which the diminishing reduction in spill oad is outweighed by the additional cost of increased storage. The Sewage Treatment Works The sewage treatment works consists of a storm tank with overflow, primary tank, oxidation ditch with clarifier and nitrifying filter with humus tank. The oxidation ditch at this site is severely overloaded, with a retention time of 2 hours at maximum flow. STOAT makes simplifying hydraulic assumptions that allow several small process units to be modelled as a large single unit. Bypasses upstream of the filter and humus tank limit the flow to these processes reducing the risk of overloading them. Figure 1 shows the STOAT model of the works. ‘The main model parameters are listed in Table 3. TABLE 3 STOAT MODEL PARAMETERS Sewage temperature (°C) 10.7 Maximum flow to treatment (I/s) 252 Maximum flow to filters (V/s) 186 Maximum flow to humus tanks (V/s) 140 Storm tank Volume (m*) 1958 Surface area (m*) 1070 Flow to works below which contents are returned (V/s) 210 Pumping rate for returns (1/s) 30 Primary Number of CSTRs * 2 Volume (m*) 1914 Surface area (m*) 740 Oxidation ditch Volume (m’) 1872 Wastage rate to primary tanks (1/s) 9 DO set point (mg/l) Ll Recycle ratio RAS:Sewage al Clarifier Surface area (m*) 1648 Depth (m) 3.27 Filter ‘Area (m?) 1236 Depth (m) 1.83 Media dimension (m) 0.25 Specific surface area of media (m"/m*) 223 Humus tank Surface area (m?) 432 * CSTR = Continuously stirred tank reactor Modelling Approach The model was first run using the MOSQITO outfall results for dry weather, until a state of equilibrium was reached, to establish a dry weather diurnal pattern for the whole works. The state of the sewage treatment works at 09:00 hours was then used as the initial state for the storm runs, which started at 09:00 hours. This procedure allowed a realistic starting position to be achieved for the subsequent STOAT model runs. The MOSQITO results for the 3 tank sizes were used as input for each STOAT run. After each run was complete, the model was restarted with dry weather flow as input, during which time the remaining storm tank contents were returned to the main flow stream. In order to allow meaningful comparisons to be made, each event was run from 09:00 hours on day 1 until the storm finished followed by dry weather flow from the end of the storm until 21:00 hours on day 2. Thus the dry weather flow input and the rainfall input were the same in each configuration: Modelling Results The combined effect of the CSO spill and the additional load from the sewage treatment works is considered for each configuration. As expected, as the size of the tank increased the volume and load entering, and leaving, the sewage treatment works increased. Table 4 summarises the duration of the STOAT runs to allow comparison of the following tables to be made. Table 5 shows the volume of flow from each source entering the river. TABLE 4 STOAT RUN DURATIONS Storage volume (m") Time (hours) 0 1500 3000 Storm 10 10 12 Dry weather flow (DWF) 26 26 24 Total time (hours) 36 36 36 nb. times reflect duration of timeseries used not duration of storm flows. TABLE 5 VOLUME ENTERING THE RIVER, a ss ‘Storage volume (m") ‘Volume (m’) 0 1500 3000 CSO spill 11019 9908 8495 STW storm tank overflow 1670.4 2059.2. 2276.6 STW final effluent (storm) 7128 79416 9560.2 STW final effluent (DWF) 8002.8 82368 © 7655.0 Total volume 27820-28146 «(27987 ‘The total volume should be equal to the dry weather flow for the 36 hours plus the volume of rainfall, and therefore should be equal for all 3 configurations. ‘The sum of the volumes calculated by the MOSQITO and STOAT models for the 3 cases differ by 1.2%. This is to be expected when modelling differing flows (downstream of the CSO) in discrete time steps. As pollutant load is the product of pollutant concentration and flow it may be assumed that differences of around 2.4% can be attributed to discretization. Tables 6 and 7 show the pollutant loads entering the river for BOD and ammoniacal nitrogen, Tespectively. TABLE 6 BOD LOAD ENTERING THE RIVER SS Storage volume (a) BOD load (kg) 0 1500 3000 CSO spill, 344 367 290 STW storm tank overflow 75.65 93.1 95.5 ‘STW final effluent (storm) 84.23 98.48 111.91 STW final effluent (DWF) 34.11 37.33 36.09 Total BOD load 738 596 534 TABLE 9 MAXIMUM AMMONIACAL NITROGEN CONCENTRATIONS ENTERING THE RIVER ‘Storage volume (m*) 0 1500 3000 CSO spill 22 19 19 STW storm tank overflow 1.52 1.32 1.2 STW final e‘fluent 9.77 10.46 10.44 Maximum concentration 9.97 10.46 10.44 The maximum BOD concentration occurs when no storage is used, consistent with some of the first foul flush spilling. The maximum BOD concentration entering the river when no storage is used is around 43% greater than the maximum BOD concentrations entering it when storage is used. The maximum ammoniacal nitrogen concentrations differ by only 7% between the 3 cases. The benefits of reducing the high BOD concentration and load to the river by introducing in-sewer storage most probably outway the effects of the slight increase in ammoniacal nitrogen. Aquatic life is most severely affected by a Grop in the dissolved oxygen levels and an increase in unionised ammonia in the river. The worst effects of the combined polluting inputs may be seen someway downstream from their various sources. A dynamic river model, such as MIKE-11, may be used to assess the effect of the various overflows and sewage treatment works outfall on the river. STOAT, therefore, includes an option to export results in the MOSQITO file format, for entry to the MIKE-11 model. The results may also be used to assess whether standards for discharges to rivers are being maintained. Conclusion The results from this application demonstrate the importance of considering urban pollution management on a catchment wide basis. In certain circumstances, introducing storage in the sewerage system to reduce CSO operation could potentially lead to an increase, rather than a decrease, in overall river pollution. STOAT can be used in conjunction with other models to help determine an optimum, cost effective method for reducing urban pollution. FUTURES STOAT is being developed in three areas. The first is to provide a direct benefit to WRe customers by using STOAT to implement, where relevant, process models for the customers to exploit WRe research. The second area is the extension of the standard process models within STOAT. New models are planned for biological aerated filters, rotating biological contactors and water treatment, The third area for development of STOAT is within the UPM programme and the field of real-time control. Various possible applications within real-time control are seen: 0 For local works operation, for example deciding when to retum sludge liquors or septic tank imports back into the main treatment stream. Built up with predictions of expected flows, this can prevent the inadvertent addition of high-strength liquors at times of peak loading. * Customising 10 take data directly from the works instruments. STOAT can then be used to carry ut basic data validation to ensure that the instrument readings are correct. This also prevents the works being operated based on incorrect data. * Using STOAT to model short-term optimization of aeration systems and pump settings. This requires the addition of blower and motor efficiency curves to the software to correctly calculate the power requirements. Q In conjunction with sewerage models, predicting the near-future flows arriving at the works. With this knowledge what-if problems can be run to evaluate preferred works operation profiles. gain with sewerage models, and preferably river models, deciding igh flows if the best operations policy is to spill from a sewer overflow, take the flow through to the works overflow or works storm tanks, or attempt to treat the flows at the sewage works. This, will require adjustable overflow settings, which is not currently a feature at many sewage works. TABLE 7 AMMONIACAL NITROGEN LOAD ENTERING THE RIVER eas eee Storage volume (m’) NH,-N load (kg) 0 1500 3000 CSO spill 93 15 61 STW storm tank overflow 2.24 2.58 2.91 STW final effluent (storm) 44.79 S121 S715 STW final effluent (DWF) 13.20 14.18 12.79 Total NH,-N load 69.5 15.5 79.0 It can be seen from Tables 6 and 7 that as the storage volume increases, the pollutant loads leaving the CSO decrease but, the loads leaving the STW storm tanks and final effluent channel increase. For this storm in this catchment (including sewer system and treatment works) the total BOD load decreases as the storage size increases but the total ammoniacal nitrogen load increases. This is due to the relative differences in magnitude of the two pollutant loads from the CSO and STW. The difference between the maximum and minimum BOD loads for the 3 configurations is 38%. Similarly the difference in ammoniacal nitrogen loads is 14%. Figure 2 shows the distribution of the BOD load for the 3 configurations. Tables 8 and 9 show the maximum concentrations from each source entering the river. TABLE 8 MAXIMUM BOD CONCENTRATIONS ENTERING THE RIVER SS Storage volume (mm) Max. BOD cone. (mg/l) 0 1500 3000 CSO spill 183 93 7 STW storm tank overflow 75.7 78.6 59.4 STW final effluent 92.0 105.8 104.1 Maximum concentration 183 105.8 104.1 BOD load - 0 cu.m storage sya aon cola svaz sinz | & oltalat ° giz i z se ‘s¥:L0% 2 a9 § nora : : 8 SLOT I . = chow & 3 i 3 i shal 3 tee 3 wai 8 oat 8 ] g suez onl Fooletfuent — — sristz ohare Figure 2 - BOD load steoz cee eeu “meeesioon ‘yOu'9001008 REFERENCES Lessard,P., Beck,B. (1989). Sciences et Techniques de I’ Eau., 22(2), 131-140. Lessard,P., Beck,B. (1988). Journal of Environmental Engineering, 114(4), 753-769. Mosey,F.E. (1983). Water Science and Technology 15(8/9), 209-232.

You might also like