Conference Agenda

Active control of pump stations in urban drainage tunnels poses a major challenge for operators, given that pump schedules must be planned pre-emptively to avert overflows while minimizing energy costs. This study derives and implements a model predictive control (MPC) scheme to determine the optimal pumping schedule for a proposed tunnel system in Austin, Texas. The proposed system consists of an 8.4 km long, 6.7 m diameter drainage tunnel that must pump water at 8.5 m³/s into a nearby river to avoid surcharging during storms. To determine the optimal pump schedule, we first develop a hydraulic model for the tunnel system based on the Saint-Venant equations for unsteady flow. We then derive and implement an MPC program that determines the optimal pump schedule while balancing between flood control and energy savings. We find that the MPC algorithm completely averts flooding during a 25-year storm event when flood control is prioritized. Conversely, when energy savings are prioritized, the MPC control strategy saves roughly 30 GJ in pump energy expenditures. Importantly, this study demonstrates a working implementation of MPC using a full physically-based model of an urban drainage system, indicating that optimal control schemes need not sacrifice model fidelity for real-time applications.