Professor Technion-Israel Institute of Technology, United States
Abstract Submission: Water distribution systems (WDS) are complex dynamic systems, requiring continuous real-time decision-making of control elements like pumps and valves to optimize several objectives. These objectives include minimizing energy costs, reducing leakage volume by minimizing redundant pressures, and optimizing water quality parameters. Traditionally, addressing these challenges involved mathematical models of the physical system. However, the complexities and inherent uncertainties in WDS dynamics often hinder the implementation and scalability of these methods. Recent advancements in data measurement, storage, and analysis have revolutionized the availability of real-time data. Coupled with the latest developments in control theory, these advancements have paved the way to improved optimization strategies for large complex systems such as WDS. The prominent advantage of data-driven methods lies in their ability to effectively bypass the need for physical system models, thus avoiding the associated computational burdens. This study explores the application of a novel Data-Enabled Predictive Control (DeePC) algorithm to optimize the real-time operation of WDS. The method employs real-time data to learn the behavior of an unknown system dynamically. It utilizes a finite set of input-output data samples (control elements settings, and measured parameters values) to derive optimal control policies. In this study, DeePC is applied to two critical control problems within WDS: pressure management and water quality optimization, demonstrating the method's capability to address diverse challenges. The DeePC method's real-time, model-independent approach, holds a significant promise for enhancing the service levels and efficiency of WDS control in an era increasingly dominated by data-driven solutions.
