Managing Loosely Coupled Networked Control Systems with External Disturbances
Abstract:
Cyber-physical system (CPS) applications for urban systems often involve sophisticated interaction and coordination between physical and computational resources with a large number of sensing and actuation devices that share information communication networks. Water or gas distribution networks, power grid transmission networks, ground/air traffic networks, and water resources networks are examples of loosely coupled networked control systems; connections among individual components of these systems or subsystems are sometimes uncontrollable. Understanding this type of control systems and being able to address technical challenges associated with their development and operation are critical to the success of these kinds of CPS applications. In our project, an interdisciplinary team of investigators combined expertise in networks, complex systems, and water resources management to study CPS applications for the Chicago Area Waterways. Through our collaboration with the Metropolitan Water Reclamation District of Greater Chicago (MWRDGC), we hope to achieve a better understanding of these loosely coupled networked control systems, and improve monitoring and control methods, and regional water quality. In 2009, in response to concerns about impaired waters stemming from nutrient loading, the USEPA issued new regulations for total ammonia, and required all wastewater treatment plants in the US to conform to the standards by 2014. As a result, wastewater utilities have been working to develop cost effective modifications for existing facilities to meet the new standards. Conventional wastewater treatment typically involves oxidation of ammonia to nitrate through a two- step nitrification process. That process can be sensitive to changes in temperature, alkalinity, dissolved oxygen, and food- to-microorganism and BOD-to-TKN ratios. Furthermore, some wastewaters contain compounds that can inhibit nitrification processes. In addition, maintaining appropriate microbial populations in the same reactor can be challenging because different organisms have different optimum growth conditions. Finally, there is an additional challenge for communities such as Chicago that have combined sewers and need to manage wet weather flows that can adversely affect nitrification processes. The Chicago Waterway System is a loosely coupled networked control system because the control connections among treatment plants and the waterways are not well coordinated. We believe that a timely and cost effective way to address these issues is to apply available technologies, such as sensors and wireless networks, to upgrade the capabilities of existing treatment plants. Such an approach can provide real-time, on-line monitoring and process control to minimize energy demands and the carbon footprint associated with nutrient control.
Major observation and conclusions from our assessment include:
We applied cluster analysis and cross-tabulation to characterize treatment plant influent, and defined five weather clusters and five composition clusters from ten years of historical data. From this information we identified nine significant combined influent scenarios that accounted for 76% of the types of influent conditions.
Storm events can have a significant impact on the treatment process. Higher flows have a greater effect on effluent quality; the worst effluent quality is typically associated with the peak effluent flow. The effluent dissolved oxygen concentration is the limiting factor that determines aeration requirements at the MWRDGC Calumet plant.
We applied an iterative, multiple linear regression method to explore how real-time ammonia and CBOD5 concentration data could be used to develop more reliable predictions of influent flow when faced with missing monitoring data.
Relative to a complete model, our reduced model retained more daily data and provided a balance between prediction performance and concerns about the potential risk of influence from missing variables.
We considered historical data, the performance of a treatment plant simulation model, and field measurements, steady- state conditions, mixing requirements, and plant response to influent perturbations from storm events. Based on our analysis, we conclude that the MWRDGC Calumet treatment plant frequently operates with excess aeration. Treatment plant response to various influent scenarios suggests that with an integrated CPS, at least a 35% decrease in aeration is possible while still satisfying all water quality effluent permit requirements.