Coupling model predictive control and rules-based control for real-time control of urban river systems

Model Predictive Control (MPC) has the potential to enhance flood control in urban river systems. However, the computational burden of online optimization hinders the application of MPC to large-scale areas. To address this, we propose a method incorporating both Rules-Based Control (RBC) and MPC based on surrogate models. MPC optimizes the operation of gates and pumping stations in the area of interest, while RBC is implemented in a specific region that does not participate in online optimization, thereby reducing calculation time. We have successfully validated this approach in coastal urban river systems in China, reducing the decision variables from 109 to 24 and narrowing the simulation area from 165 km2 to 15 km2. During the Typhoon Lupit rainfall test, the total computation time and the highest online optimization time were only 33.9 % and 38.1 % of the full MPC system, respectively. Additionally, the method demonstrated impressive results, including a significant 94.62 % reduction in flood volume and a remarkable 96.5 % decrease in flood duration. These findings demonstrate that our method effectively addresses the challenge of long computation times for online optimization in urban river systems and presents a novel approach for implementing urban flood mitigation strategies.