More than 700 cities in the U.S. have sewer systems that combine sanitary and storm water flows in the same system. During rain storms, wastewater flows can easily overload these combined sewer systems, thereby causing operators to dump the excess water into the nearest river or stream. The discharge is called a combined sewer overflow (CSO) event [1]. The discharged water is highly impacted with biological and chemical contaminants, thereby creating a major environmental and public health hazard. Under the provisions of the 1972 clean water act, the environmental protection agency (EPA) has begun fining municipalities for CSO events. These fines are substantial, sometimes running into the tens of millions of dollars. Municipalities have therefore begun looking for cost effective ways of reducing the frequency of CSO events.
The straightforward solution to the CSO problem is to enhance existing sewer infrastructure by separating storm and sanitary flows. Other solutions involve increasing the capacity of the wastewater treatment plant (WWTP) or building large off-line storage reservoirs. All of these options are extremely expensive and highly disruptive to the community.
Another solution uses the excess storage capacity in a city’s sewer to temporarily store water during a storm. This option is referred to as in-line storage. The economical and reliable control of CSO events through in-line storage requires realtime monitoring and control.
Current approaches to real-time monitoring and control of sewer systems do not scale well. Sensor data is usually collected by a single computer over a Supervisory Control and Data Acquisition (SCADA) network. This computer determines the control action and distributes it back to the system through the SCADA network. It takes time to gather all of the sensor data and the delay introduced by gathering this data will also limit the rate at which control commands can be fed back to the system. Due to these delays, the control must be computed using complex simulation models of the sewer system. The entire control problem is therefore viewed as a large-scale nonlinear optimal control problem [3] which can be addressed using linear quadratic approaches [3] or model predictive control methods [4] [5]. These controllers are always implemented in a centralized fashion for very high order plants. The system model is highly nonlinear with a great amount of uncertainty. As a result, centralized control of sewer systems tends to be complex, computationally intensive, and certainly is not robust to modeling error. All of these factors conspire to limit the scalability of centralized approaches to sewer flow control.
An alternative ”distributed” approach to CSO control was presented by Ruggaber et al [6]. This case study used an embedded network of microprocessor controlled sensors and actuators to control CSO events. The network used a simple local feedback scheme to control a stretch of sewer system fed by a 1500 foot wide by 3.2 mile long corridor. In its first month of service the network prevented a 2 million gallon CSO event. The cost of the deployed network was around $25,000, which was half of what it would have cost using existing SCADA network technologies.
The sensor-actuator network used by Ruggaber et al, therefore appeared to provide a cost-effective solution for controlling CSO events. The control used in that study was a simple switching law. A more sophisticated distributed control strategy was developed by Wan et al. [7] . This controller was a distributed control scheme that would be implemented on the WSAN used by Ruggaber et al. for a city the size of South Bend Indiana. Wan et al. were able to establish the optimality of their distributed scheme and through simulation studies were able to show that such an approach could reduce CSO overflows by 20 % over the existing passive strategies in use. On the basis of this early work, the control strategy developed by Wan et al. is being proposed for implementation on the metropolitan scale CSOnet for South Bend Indiana.