Hardware-in-the-Loop Validation of Successive Linearization-Based Predictive Control for Optimal Water System Operation

Abstract Submission: Model predictive control (MPC) in water distribution systems (WDSs) is gaining recognition for its ability to handle physical constraints, optimize various operational objectives, and predict the system's future behavior. As the WDS's complexity and nonlinearity increase, the optimization problem becomes more intricate and resource-intensive, affecting the computational cost and the optimality of the control solution. Successive linearization MPC (SL-MPC) addresses this by linearizing the nonlinear dynamic model at each time step, providing faster and feasible control. This paper marks the first exploration of experimental validation of the SL-MPC framework via a hardware-in-the-loop (HIL) technique integrated with supervisory control and data acquisition (SCADA), representing real-world system applications. The control system is applied to a quadruple tank system (QTS), incorporating a reservoir, two pumps, interconnected pipes, valves, and four tanks equipped with pressure sensors. HIL enables real-time testing of control algorithms by interfacing the control software with physical hardware components, offering a realistic and cost-effective platform for rigorous control assessment without relying on a full-scale WDS. When combined with SCADA, a comprehensive verification of both local control and supervisory functions is facilitated, enabling interactive real-time monitoring and system visualization in response to control signals. A series of HIL case studies are presented, including time-varying set-point tracking for the QTS and a comparison between SL-MPC and a state-of-the-art controller, proportional integrator (PI). The findings indicate the potential use of the SL-MPC framework for large-scale WDSs and demonstrate the practical advantages of HIL in verifying advanced control strategies, offering scalable and low-cost WDS testing.

Learning Objectives/Expected Outcome (Optional) :