EPA/600/JA-01027
                                                            2001
HED: Get Real! D

DECK: Implementing real-time control schemes offers combinedD

sewer overflow control for complex urban collection systemsD



Richard Field, Elise Villeneuve, Mary K. Stinson, NathalieD

Jolicoeur, Martin Pleau,  and Pierre LavalleeD
     Combined sewer overflow  (CSO) is a significant source

of pollution in receiving waters. However, implementing a

real-time control scheme operates automatic regulators more

efficiently to maximize a collection system's storage,

treatment, and transport capacities, reducing the volume and

number of CSOs.  Real-time control schemes are being used to

manage complex urban collection systems around the world,

including a demonstration study in Canada for the Quebec

Urban Community  (QUC)  collection system. Funded by the U.S.

Environmental Protection Agency  (under a contract to the

Office of Research and Development)  to assess the use of
                              ID

-------
real-time control schemes, the QUC study evaluated the




effectiveness of three schemes in managing CSOs.









THE QUEBEC URBAN COMMUNITY TERRITORY DEMONSTRATION SITE




     Located on the North shore of the St-Lawrence River,




the QUC territory covers 200 mi2  (500 km2), has a population




of 500 000,  and is composed of an Eastern and Western




catchment. The QUC study team's evaluation of real-time




control schemes focuses on the Western catchment.




     The Western catchment covers 65% of the QUC territory,




with close to 50% of the total population  (230 000).




Wastewater is conveyed through 41 mi  (66 km) of interceptor




pipes to a 82-mgd (310 000-m3/d) wastewater treatment plant




(see Figure 1, p. xx). The collection system has three main




interceptor branches and two tunnels that together provide




approximately 3.4 MG (13 000 m3) of inline storage. The




overflows of the western collection system represent 528 MG




(2 million m3).




     Nine of the 22 regulators have significant overflows




that empty into the St-Charles and St-Lawrence rivers. The




Dijon, Jones, and Suete CSO structures and the Affluent, and




Versant-Sud tunnel regulators overflow into the St-Lawrence




River and the Hopital,  Lessard-Durand, Talus, and Myrand CSO




structures overflow into the St-Charles River.
                              2D

-------
     The QUC's long-term CSO control plan is to maximize the




Western catchment's intercepted flow and existing




facilities, two inline storage tunnels, and the treatment




plant. The long-term plan includes implementing an Optimal




Global Predictive (OGP)  real-time control scheme in the




entire system and constructing offline storage facilities




and is projected to control more than 85% of CSOs and cost a




total of $107 million, 37% less than before implementing the




OGP scheme.
                              3D

-------
Real-time Control Scheme Simulations




     The QUC study team evaluated three real-time control




schemes using 32 real rainfall events ranging from very




small events to a large once-in-5-year event, as well as




back-to-back rainfalls between July 1 and August 28, 1988.




Five raingauges collected data to represent, in part, the




Western territory's rainfall heterogeneity. The data then




were translated into combined collection system flow rates,




which were fed to a custom-built, nonlinear hydraulic model




Using the model and simulation software, a total of 128




simulations were carried out to observe the performance of




the three different control schemes. The control schemes




were evaluated for CSO volumes, number of CSO events,




surcharge occurrence, treatment plant utilization,  and




inline storage capacity.
                              4D

-------
Real-time Control Schemes




     The study team investigated  (1) Local Reactive Control




(LRC Type 1) that operates collection system gates at fixed




flow set points on the intercepted flow, requiring local




site control;  (2) Local Reactive Control  (LRC Type 2), which




works similar to Type 1,  except it operates the gates at




both fixed and variable set points in respect to flow




capacities located at some specific pipes; and  (3)  OGP that




operates the gates at optimal variable set points proactive




to actual rainfall conditions, which predicts flow 2 hours




in advance using rainguages and flow and rainfall prediction




models.




     Implementing the LRC Type 1 scheme can be as simple as




employing a mechanical device to open or close a system gate




while the Type 2 scheme is more complex, similar to the OGP




scheme.  Both the LRC Type 2 scheme and the OGP scheme




require more instrumentation and equipment; however, the OGP




scheme differs by using a central decision-making system,




prediction models, and other more sophisticated programs and




equipment.




     Selecting a real-time control scheme depends on the




architecture of the collection system and the environmental




objectives pursued. Collection systems with small storage




capacities, few flow control devices, and restrictive flow
                              5D

-------
constraints can be properly managed with relatively simple




control schemes but more complex systems require a global




control approach. Selecting a real-time control scheme not




only depends on performance, it also depends on criteria




such as implementation and process control, capital cost,




and operations and maintenance costs.









Control Objectives




     Within the Western catchment's existing collection




system (currently without offline retention tanks), the




selected real-time control scheme must:





     • Reduce  CSO  frequencies  and volume  as much  as





possible during operational season activities  (from May 15




to September 15)  to meet water quality levels for contact




with the St-Charles and St-Lawrence rivers;





     • Eliminate  surcharge  flow  caused by flooding from





private connections along the inceptor at a setting of 95%




of its total capacity;





     • Allow variable flow  set points  to  maximize  the





Western treatment plant's capacity, which fluctuates with




the St-Lawrence River tide;  and





     • Use  the Western system's  two  major inline  storage





tunnels to maximum capacity and ensure no premature
                              6D

-------
overflows occur while residual storage capacity is




available.









Overall Performance




     In comparing the three control schemes with one




another, the OGP scheme has the lowest CSO volumes and




number of CSO events (see Figures 2 and 3, p. xx). Because




it can constantly readjust its control set points according




to updated field information,  this control scheme is the




most efficient to control and minimizes the surcharge flow




in the system (see Figure 4,  p. xx).  The OGP scheme also




permits programming more sensitive overflow sites as




priorities and allows the system to constantly adapt to




protect these sites. In fact,  the more complex the





collection system — number of flow paths  and storage





options —  the better the OGP  scheme  performs.





     The difference in total  CSO volume between the OGP




scheme and LRC Type 2 is relatively small compared to the




total CSO volume recorded with the other two control




methods. The LRC Type 2 scheme did not eliminate surcharges




and is not flexible enough to properly manage future offline




storage facilities. The LRC Type 2 scheme is more suitable




for controlling relatively simple systems that accept a
                              7D

-------
certain amount of surcharge. At this time, without offline




storage tanks, the OGP scheme behaves as a flow management




scheme that conveys the maximum amount of water to the




Western treatment plant, within the flow constraints.




     Using an August 27, 1988 rain event as an example, the




inflows to the Western treatment plant without a real-time




control scheme and under OGP control show that no overflow




occurs at the plant with the OGP scheme  (see Figure 5, p.




xx).  However, without a real-time control scheme, a 0.45-MG




(1700-m3)  overflow occurred.  The OGP scheme  also conveyed 78




MG (295 000 m3)  of combined wastewater to the plant while




operating the system with no real-time control scheme only




conveyed 72 MG  (271 000 m3)  and allowed a 6-Mg (22  700 m3)




overflow at the plant. Furthermore, without a real-time




control scheme, the Versant-Sud tunnel was used only as a




conveyance system, whereas under the OGP scheme, the tunnel




also was used for storage  [up to 2-MG  (8000-m3) ] .
                              8D

-------
Implementation and Process Control




     Implementing either LRC scheme poses more operations




and maintenance concerns, depending on the quality and




quantity of measurement and control devices installed. To




maintain a prescribed flow set point,  flow routines must be




programmed and calibrated, and controllers, such as





Proportional Integrative Derivatives  (PIDs) — mathematical





equations used to adjust the position of the system gates —





need to be implemented and properly tuned. Downgraded




management modes must be defined and implemented at the




local control stations to address equipment breakdowns or




other system anomalies and should include predefined flow




and gate opening set points for every kind of foreseeable




failure or breakdown. Finally, a telecommunication system




and a central supervisory control station are recommended to




monitor the performance of the control scheme.




     The difficulty of implementing the telecommunication




system varies with the topography of the territory covered




by the collection system. For the Western network, the land




is relatively flat,  ideal for using a radio




telecommunication system. Moreover, the fewer local control




stations, the less data traffic to interfere with




telecommunications.
                              9D

-------
     Implementing the OGP control scheme requires a more




sophisticated level of process control. The difficulties




encountered are similar to those described for both LRC




schemes. However, design parameters require determining




variable measurements and accuracy of the hydraulic model,




transmission distortion of control signals, meteorological




predictions, and flow set points using optimization (the




equivalent to an "intelligent" decision-making machine) and




filtering algorithms  (such as averaging or exponential




computation) and nonlinear programming. In addition, the




implementation of a central control station is more complex.




An optimal control problem has to be setup and solved in




real-time using an optimization algorithm. A meteorological




forecasting model, calibrated with raingauge measurements,




may be needed to guarantee good performance. If the




forecasting algorithm relies on radar images,  the




availability of these images in real-time must be




considered.









Capital Cost




     The capital cost of implementing a real-time control




scheme depends on the quality and quantity of control and




measurement devices required for a successful




implementation, as well as the models and the algorithms
                             ion

-------
needed to compute the flow set points. A preliminary study




of the Western collection system shows that implementing the




OGP scheme costs approximately $4 million  (less than 4% the




total cost of QUC's long-term CSO plan),  the LRC Type 2




costs approximately $2.5 million, and the LRC Type 1 costs




approximately $1.5 million.









Operation and Maintenance Costs




     For the Western network, the real-time control schemes





are in operation only during the regulated period — May 15





through September 15, meaning there are no maintenance costs




for a significant period of the year.




     Operation and maintenance costs depend on the




sophistication of the implemented control scheme (the number




of control and measurement devices, as well as the




geographical characteristics of the collection system).




Implementing any one of the three real-time control schemes




can be a relatively inexpensive solution compared to




conventional alternatives. In fact, in the QUC study each




scheme represents less then 4% of the total cost for




complying with long term CSO control regulations. Operating




the mobile actuators, telecommunication systems, and




supervisory systems generate electricity costs and certain




models require regular purchases, such as radar images if







                             11D

-------
using a meteorological forecasting model. However,  the OGP




scheme allows for additional control objectives to minimize




electricity costs for pumping and treatment.




     Operations and maintenance considerations for




implementing any real-time control scheme includes weekly




cleaning of sensors, monthly testing of programmable logic




controllers and personal computers in downgraded mode, and




regular mechanical maintenance of gates and actuators. For




implementing the OGP scheme, additional operations and




maintenance considerations include calibrating and




validating meteorological forecasting model every 3 months.




Quality control must be performed on the database processing




archives monthly and after each rainfall event. Quality




control also must be performed on the collection system




configuration every 3 months and after any modifications.




The hydraulic models must be calibrated yearly, and




statistics and reports on performance and default conditions




must be compiled monthly and after each rainfall event. The




decision-making system, control objectives, and global and




local priorities also must be verified and adjusted monthly




and after each rainfall event. Constraints included in the




non-linear programming algorithm must be verified and




adjusted monthly.
                             12D

-------
Real Results




     The QUC example demonstrates the potential of real-time




control schemes in maximizing the capacity of collection




systems and reducing CSOs.  Even with a relatively simple




system with no offline storage to manage, the real-time




control schemes evaluated in the QUC reduced CSO volumes by




24% to 47%, representative of potential performance in most




collection systems. However, real-time control schemes




should be selected depending on a collection system's




configuration and the control and operational objectives




specified by the utility authority.









     Richard Field is the project leader and a senior




environmental engineer and Mary K. Stinson is a physical




scientist for the U.S. Environmental Protection Agency, Wet-




weather Flow Research Program,. Urban Watershed Management




branch, Water Supply Water Resources division, National Risk




Management Research Laboratory  (Edison, NJ).  Elise




Villeneuve is a project director at BPR Consultants




(Montreal, P.Q., Canada). Nathalie Jolicoeur and Martin




Pleau are project engineers and Pierre Lavallee is the




senior project director and executive vice president of BPR




Consultants  (Quebec, P.Q., Canada).
                             13D

-------