Model predictive control of stormwater basins coupled with real-time data assimilation enhances flood and pollution control under uncertainty

Abstract

Smart stormwater systems equipped with real-time controls are transforming urban drainage management by enhancing the flood control and water treatment potential of previously static infrastructure. Real-time control of detention basins, for instance, has been shown to improve contaminant removal by increasing hydraulic retention times while also reducing downstream flood risk. However, to date, few studies have explored optimal real-time control strategies for achieving both water quality and flood control targets. This study advances a new model-predictive control (MPC) algorithm for stormwater detention ponds that determines the outlet valve control schedule needed to maximize pollutant removal and minimize flooding using forecasts of the incoming pollutograph and hydrograph. We illustrate that, compared to rule-based controls, MPC more effectively prevents overflows, reduces peak discharges, improves water quality, and adapts to changing hydrologic inputs. Moreover, when paired with an online data assimilation scheme based on Extended Kalman Filtering (EKF), we find that MPC is robust to uncertainty in both pollutograph forecasts and water quality measurements. By providing an integrated control strategy that optimizes both water quality and quantity goals while remaining robust to uncertainty in hydrologic and pollutant dynamics, our study paves the way for real-world smart stormwater systems that will achieve improved flood and nonpoint source pollution management.