Data Enabled Predictive Control for Water Distribution Systems Optimization

Abstract

Recent developments in control theory, coupled with the growing availability of real-time data, have paved the way for improved data-driven control methodologies. This study explores the application of the Data-Enabled Predictive Control (DeePC) algorithm to optimize the operation of water distribution systems (WDS). WDS are characterized by inherent uncertainties and complex nonlinear dynamics. Hence, classic control strategies involving physical model-based or state-space methods are often difficult to implement and scale. The DeePC method suggests a paradigm shift by utilizing a data-driven approach. The technique employs a finite set of input-output samples (control settings and measured data) to learn an unknown system's behavior and derive optimal policies, effectively bypassing the need for an explicit mathematical model of the system. In this study, DeePC is applied to two WDS control applications of pressure management and chlorine disinfection scheduling, demonstrating superior performance compared to standard control strategies.