Development and commissioning of decision support tools for ...

Development and commissioning of decision support tools for Sewerage Management G. Manic*, C. Printemps*, M. Zug**, C. Lemoine* * Anjou Recherche, R&D Center of Veolia Water, Chemin de la Digue, 78603 Maisons Laffitte, France ** Compagnie Générale des Eaux, Région Ouest, 3 rue Marcel Sembat, 44925 Nantes, France

Abstract Managing sewerage systems is a highly complex task due to the dynamic nature of the facilities. Their performance strongly depends on the know-how applied by the operators. In order to define optimal operational settings, two decision support tools based on mathematical models have been developed. Moreover, easy-to-use interfaces have been created as well, aiding operators who presumably don’t have the necessary skills to use modelling software. The two developed programs simulate the behaviour of both Wastewater Treatment Plants (WWTP) and sewer network systems, respectively. They have essentially the same structure, including raw data management and statistical analysis, a simulation layer using the Application Programming Interface of the applied software and a layer responsible for the representation of the obtained results. Four user modes are provided in the two software including the simulation of historical data using the applied and novel operational settings, as well as modes concerning prediction of possible operation periods and updates. Concerning the WWTP software, it was successfully installed in Nantes (France) in June, 2004. Moreover, the one managing sewer networks has been deployed in Saint-Malo (France), in January, 2005. This paper presents the structure of the developed software and the first results obtained during the commissioning phase.

Keywords Decision support tool, Waste water treatment plant, Sewer network, Optimization

INTRODUCTION Waste water management has become a highly complex task and requires huge know-how from the operators to fulfil the quality constraints on performances. This is true for waste water treatment plants (WWTP), to optimize discharge concentrations by minimizing energy costs, but also for sewer network management, to reduce overflow to the natural environment by optimizing rainfall events management. In order to help the operators to control and optimize the waste water facilities, Anjou Recherche developed models to better understand the biological and physical mechanisms. Two pilot sites had been chosen: Saint Malo, for sewer network modelling and Nantes Tougas for WWTP modelling. Two commercial simulators were used to develop these tools: Infoworks-CSTM (sewer networks modelling) from Wallingford Software Ltd © and WEST® (WWTP modelling) from Hemmis ©. One of the major difficulties of modelling tools is that the field-operators mostly don’t have the time or the computer-science skills to handle there models, so this mainly remains on expert level (R&D or engineering offices). The aim of this study is to develop decision support tools which use the mathematical module of commercial software, automatically setup input data, offer a user friendly interface and functionalities needed by operators. This article presents the

parallel methodology used for the development, the different user modes and the operational results for both WWTP and sewer network management. MODEL DESCRIPTION WWTP modelling (Nantes-Tougas) with WEST© (Printemps et al, 2004) The WWTP of Nantes Tougas is an activated sludge treatment plant sized to treat a 600,000 population equivalent flow collected in a combined sewer network. The plant receives a wastewater inflow of about 70,000 m3/day. A project was started in 2002 to install on site an activated sludge model and adapt it to help the operators managing the WWTP (Printemps et al, 2003). The model was calibrated and validated over a for months period. The model was created with the mathematical model of the software WEST© from Hemmis© (Hemmis, 2002). The original model was extended by adding new modules. The activated sludge part, for example, was described using the extended ASM1 model developed by Anjou Recherche (Lesouëf 1990). Sewer network modelling (Saint Malo) with INFOWORKS© CS The model used for the development of a decision support tool for sewer network management was the one developed for the city of Saint Malo, a tourist resort on the French Atlantic coast. The model building is a part of a project of integrated waste water system management to improve bath water quality by reducing pollution overflow in case of rainfall events. The sewer network of the city is a combined network presenting two majors characteristics: a strong season-depending population (from 50.000 to 120.000 inhabitants) and a big tidal influence due to the proximity of the sea. The network has twelve main subcatchments, covers a surface of about 1600 ha and has a imperviousness coefficient of 0.51. The model was built using Infoworks© CS, from Wallingford Software Ltd ©. The model was calibrated and validated over a period of six and five months, respectively (Gogien et al, 2002). Decision support tool architecture Both Infoworks© CS and WEST© contain an Application Programming Interface (API). An API allowed a link between the simulation software and other programs, like Microsoft Excel© or Microsoft Visual Basic©. Thanks to an API, it is possible to set simulation input data, run a simulation and access or export the results with another user interface than the simulator. The architecture of the decision support tool uses this software possibility and is divided into three layers (Figure 1). A first layer will link the process simulator with the on-site instrumentation through Supervisory Control and Data Acquisition system (SCADA) as well as data logger, or self measurement data to collect, filter and treat raw data. A second layer will use the API of the simulation software to handle simulations and calculations aspects. Then a third layer will display results through graphs or reports, as shown in Figure 1.

Figure 1 Architecture of the decision support tool

The user interface was developed in Microsoft Visual Basic©. This programming language was chosen because of its good compatibility with the software API and its visual interface possibilities. The ActivX controls of “ultrapack” from IOComp© software and Microsoft Office XP© were used to edit simulation reports and generate graphs. These tools are still prototype versions, but almost 80% of the source code be reused to develop the tool on a new site. The 20% left are mostly user interface modifications, adaptation on model characteristic and implementation of specific functionalities needed by the operator. SET UP OF INPUT DATA The first crucial step of modelling is to ensure valid input data in order to perform correct simulations. The format of input data is a critical phase to run valid simulation and have a good decision support tool. The automation of this step was pushed as far as possible in order to reduce errors. The user interface was made so that only a few mouse clicks are enough to collect data from different sources, filter them and generate input data for the simulator. Raw data sources Raw data are delivered by several sources. The tool has to automatically pick up data, filter them and spread them into the model. Three data sources exist: SCADA systems, self measurement monitoring values and data-loggers. First, the data from SCADA systems (Nantes, St Malo) are collected. In case of WWTP models, it gives the major operational settings of the facility. For the sewer network, the SCADA system delivers state values of the network, like tide levels or valve position. Then, self monitoring

values are used. These data are mainly offline measurement still disconnected from the SCADA system (e.g. inflow concentration value – NH4, NO3, COD … – of WWTP). These values are recorded in a standard monthly file. Finally, raw data are delivered by data logger in specific text-files (e.g. Rainfall data, for sewer network simulation). Automatic Control and Filtering All data coming from different sources are statistically filtered and controlled. Some of them can be wrong or missing. These failures can have several causes: measurement failure, loss of data, unrealistic values … Statistical tools had been then developed for satisfying two objectives: • Validate the data: this is done by minimal and maximal values, standard deviation analysis or moving average; • Complete data series if information is missing using a statistical forecast model to suggest alternative values. This step is crucial because invalid input data of a model leads to low quality simulation results. The automation of this phase avoids a fastidious file review and brings great time saving. It also standardizes the way of filtering data. Data validation by operator The final validation is impossible without an expert review. A graphic interface was added to help the final user to verify the data before a final validation. The tool prompts the user if data were changed or estimated. The operator can then decide if the data are correct enough, and also choose eventually a simulation strategy. For example, in case of a WWTP simulation, unrealistic data on airflow rate for the activated sludge aeration are often recorded. Measurement failures on airflow rate are detected by the filtering module. Two simulation strategies are possible, by using airflow rate or by using dissolved oxygen concentration. DECISION SUPPORT TOOL FUNCTIONALITIES Four user modes have been defined for the tool: • Mode 1 : Simulations using historical data • Mode 2: Mode 1 + new operational settings • Mode 3: Forecast • Mode 4: Update Mode 1 – simulation over historical data This mode enables the simulation of a past period using archived data. The results are automatically compared with the values of field-measurements, coming from self monitoring analysis or from SCADA (or supervisory) system. It allows the operator to verify the model robustness over a long simulation period and determine the final state of the facility, which will be used for forecasts and optimization in user modes 2 and 3. To launch a simulation, it is possible to define a specific initial state or use a standard one (see Figure 2 and Figure 3 – user interfaces for sewer system and WWTP tools, respectively). The results of simulation are displayed through graphs or performances tables. Automatically generating reports at specific time intervals is one of the first needs fulfilled by the decision support tool.

Figure 2 Simulation configuration - sewer network simulation

Figure 3 Simulation configuration - WWTP simulation

Mode 2 – Mode 1 + new set points This mode allows the simulation of past periods using new operational settings. The aim of this mode is to try alternative set points configurations in order to optimize the management by improving system performances and / or energy consumption. Another goal of this mode is to study the sensibility of the parameters of the process without testing it full scale, which gives to the decision support tool a training aspect. In case of the tool for sewer network management, the user can define new real time control rules and determine the performances in term of waste water overflow and intercepted pollution (Figure 4). For the WWTP simulation tool, it is possible to change airflow rates, return and waste activated sludge (RAS and WAS) and to evaluate the effects on the energy consumption or pollutants removal (Figure 5).

Figure 4 Real time management comparison - sewer network simulation

Figure 5 Changing operational set points - air flow rate and sludge extraction - WWTP simulation

Before using the mode 2, it is important to first validate the model over the time-period by simulating it on mode 1, and compare the simulation results with the measured data.

Mode 3 – Forecasts This mode gives the user the possibility to forecast and anticipate the behaviour of the system over the next days. Thanks to the functionalities in mode 1, the actual state is calculated using recent historical values. The user can then define the following day’s configuration and observe plant behaviour by choosing specific operational set points. In the case of WWTP simulations, a small statistical module links the rainfall data and the inflow volume and concentration (see Figure 6). The relationship between rainfall events and inflow characteristics was determined statistically using 3-months data. Further developments of the tool will improve this model. For the sewer network simulation, a design rainfall generator was created (Figure 7) to simulate the network behaviour using a user-defined initial state and estimate the amount of intercepted flow and overflow-concentration.

Figure 6 WWTP inflow prediction

Figure 7 Sewer network - Design Rainfall

Mode 4 – Update Once the model is calibrated and validated with given time periods and operating conditions, it will reproduce the behaviour of the process in a specific range. If the process is changed, for example a new aeration system in the WWTP or an alternative regulation system for sewer network, the model won’t be able to reproduce correctly the behaviour of the system. This mode allows the user to change parameters of the model, to readapt it, if needed, to a new configuration. However, due to the complexity of the task, the access to this mode remains at the expert level, following methodology of (Oborzynska, 2000). RESULTS On site installation The tools have been installed partially on site. On the one hand, the sewer network modelling tool remains at Anjou Recherche for further development. On site implementation will be done at the end of the year 2004 after finishing the development of the data treatment layer. However, specific studies of different management scenarii have been performed using mode 2 of the tool. On the other hand, the WWTP decision support tool has been installed on site for several months in the WWTP of Nantes Tougas. After a short training period, the operators of the plant now use weekly the tool. In order to meet their needs, meetings are organized frequently for further implementations of the tool by adding new functionalities.

Input data analysis Data analysis over one year of operation on the WWTP of Nantes Tougas allowed the detection of several measurement failures. The most recurrent detected error concerns the estimation of air flow rate on a couple of treatment trains of the WWTP, which leads to abnormal simulation results. Thanks to the statistical module of the tool, an alternative simulation was proposed which improves the qualities of the estimations. A comparison between simulations with erroneous and valid data can be seen in Figure 8 and 9. In the first case, the model (plotted in curve) cannot reproduce the real behaviour of the plant (plotted in histogram). In the other hand, using corrected input data, the model is valid. The discrepancy observed between two simulations is due to a error in the measurement of air flow rate of one of the treatment trains of the WWTP (flow set to zero causing N-NH4 peaks at the outlet).

Figure 8 N-NH4 & N-NO3 concentration – simulation made with abnormal input data

Figure 9 N-NH4 & N-NO3 concentration -– simulation made with valid input data

Process management improvement One of the most valuable functionalities of these tools is the mode 2, which offers a return on experimentation by giving the possibility to simulate past periods by changing tuning points of the process. Thanks to the tool, a study was made to see the effects of an airflow rate reduction on WWTP performances in term of ammonia and nitrate concentrations in the effluent. After verifying that the model was valid on the period (the simulation following the original tuning fitted the measurement); a simulation was run with modified air flow rates (see Figure 10). Results showed that lower air flow rates (up to 20% less) leads to strong energy costs and better denitrification, while ammonia concentration at the outlet remains under acceptable limit (< 5 mg/L). For the sewer network tool, an alternative real-time management based on effluent concentration was tested and compared to initial management. Results are showed on Figure 11. The graphs represent the fraction of intercepted (blue) and discharged (red) effluent during rainfall events for both initial (left) and modified (right) real time management. The proposed real time control allows a clear reduction of pollution discharge in term of TSS and COD concentration (respectively 14% and 10.1% lower).

Figure 10 WWTP - Air flow rate reduction

Figure 11 Sewer Network - Optimization of waste water discharge

CONCLUSIONS AND OUTLOOK Decision support tools based on mathematical models of commercial modelling software offer the possibility for the field-operators to better understand their facilities and optimize performances. The successful transfer of the software to operators allows further developments. The first development will be the linkage between SCADA-system and the software, to get online values. Then, possibilities of linking two models - WWTP and sewer networks - will be studied, in order to get a full integrated sewerage system. ACKNOWLEDGMENT The authors would like to thank the urban communities of Nantes and St-Malo for their kindly support, and the managers of Nantes-Tougas (N. Chatal, M. Pouliquen J.Y. Foquereau) for their fruitful contribution. REFERENCES Gogien F., M. Zug, J. Le-Luherne, and G. Jamet. (2004). Pollution modelling on the site of Saint-Malo : an interesting tool for sewer systems optimization. In Proc. Fifth international conference Novatech', pp.367-374. Lyon, France. Hemmis, N.V. (2002). WEST® models guide v.3: On-line documentation. Hemmis nv, Kortrijk,Belgium. Lesouëf, A. (1990). SIMBAD: un modèle mathématique pour systèmes de boues activées. Techniques Sciences Méthodes - L'Eau, 85(7-8), pp 371-378. Oborzynska, A., Oborzynski, K., Urbaniak, A, (2003) Component-oriented approach for management and control of wastewater treatment plant, In Proc. IV-th international conference “Water Supply and Water quality., September 11-13, Krakow, Poland. Printemps, C., Baudin, A., Dormoy, T., Vanrolleghem, P.A., Zug, M. (2003). Optimisation of a large WWTP thanks to mathematical modelling. In Proc. 9th IWA Specialised Conference on Design, Operation and Economics of Large Wastewater Treatment Plants. September 1-4, Prague, Czech Republic. Printemps, C., Manic, G., Zug, M. (2004), An Operational Modelling Tool For The WWTP Management, In Proc. 4th IWA World Water Congress, September 19-24, Marrakech, Morocco.