Coupled, Data-Driven, and Real-Time Modeling and Control of Sewer Systems and Water Resource Recovery Facilities

Abstract

Within the urban water cycle, the challenges posed in the operation of combined sewer systems include changing storms, evolving regulations, and impacts to environmental health. While building bigger infrastructure is one way to solve issues such as sewer overflows, budgetary constraints and increasing stresses to the system, such as climate change, limit the feasibility of this option for many communities and utilities. One alternative is posed by an increasing availability of sensors and data algorithms. Rather than building bigger, the use of real-time data and remote actuation provides a new avenue to autonomously adapt performance of the entire existing system.

While promising, there are outstanding knowledge gaps that must be closed to bring the idea of smart wastewater systems to fruition. 1.) Sewer systems are highly dynamic and spatially heterogeneous. Thus a static, one-size-fits-all modeling approach will not accurately reflect the real-world system. This dissertation addresses this by presenting a data-driven toolchain that learns from historical sensor measurements to estimate current and future combined sewer conditions. By evaluating this toolchain on sensor data collected across the Detroit combined sewer network, it is discovered that wastewater and stormwater flow components exhibit distinct spatial and temporal variation, underscoring the importance of flexible re-calibration using the most relevant window of data. 2.) The efficacy and feasibility of real-time control across the sewershed poses a number of challenges. In particular, objectives for control across the scale of a city often force trade-offs between flood reduction and water quality; without informing control decisions based on these trade-offs, unintended consequences will affect performance across the system. To address this challenge, this dissertation introduces a real-time control algorithm to balance loads across distributed sewer assets and equalize combined sewer flow. The algorithm is evaluated in a simulated subsection of the Detroit combined sewer network. Trade-offs between flow and water quality objectives are evaluated to inform algorithm parameterization and considerations toward implementation. 3.) While the individual control of either sewer networks or water resource recovery facilities (WRRFs) has been explored separately, the opportunity to link these system components must consider the impact that sewer control has on WRRF operation and performance. By focusing on chemical phosphorus treatment, this dissertation quantifies the impact that WRRF influent dynamics and chemical addition has on treatment efficacy and efficiency. Namely, leveraging these two strategies together, phosphorus treatment is maintained or even improved, while chemical consumption is reduced. These findings exemplify benefits that can be accomplished by coupling the control and operation of system-wide assets.