Loading...
Data-driven process monitoring and control in municipal wastewater treatment
Newhart, Kathryn B.
Newhart, Kathryn B.
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2020
Date Submitted
Keywords
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
The American Society of Civil Engineers estimates that by 2040, water and wastewater infrastructure (e.g., treatment facilities, conveyance systems) in the United States will require $144 billion to repair, replace, and upgrade. As water demand increases and stringent nutrient regulations are implemented, facilities will be forced to intensify existing operations or face costly process additions. Data-driven process monitoring and control is an underdeveloped paradigm in the treatment industry that could help address the growing infrastructure investment gap by improving the accuracy and precision with which water and wastewater are treated. Conventional wastewater treatment plants (WWTP) use relatively simple control systems compared to other process control industries (e.g., energy, manufacturing). “Normal” is demarcated by operator-determined thresholds on individual process variables; the upper and lower limits span a sufficiently large range to allow for some operational and environmental variability to water quality and quantity. Often these thresholds are so generous, process efficiency, and in some cases quality, is negatively impacted before an operator can respond to a fault or failure. Automated process adjustments are also limited by simple control logic, and process models lack the accuracy for real-time control. This frequently results in excessive energy and chemical use to safely operate in the event of a rapid and unexpected process upset. To improve the accuracy of monitoring and control, control strategies must consider the real-time multivariate and dynamic features of water treatment. This dissertation investigates multiple statistical and machine learning approaches to address control challenges faced by full-scale WWTP. First, a literature review details and compares state-of-the-art and state-of-the-industry WWTP monitoring and control methods. Second, multivariate statistical process control methods and adaptations are compared to identify faults in a decentralized WWTP. Third, a full-scale biological treatment process is (i) quantitatively assessed for stability and (ii) modeled to forecast ammonia for advanced control. Fourth, disinfection performance is modeled to adapt to changing water quality at a full-scale WWTP. Each chapter considers the realities of designing real-time monitoring and control using nonstationary, autocorrelated, and missing data, and solutions are proposed that use traditional and novel data science tools.
Associated Publications
Rights
Copyright of the original work is retained by the author.