High‐ P erformance I ntegrated C ontrol of water quality and quantity in urban water reservoirs

This paper contributes a novel High‐Performance Integrated Control framework to support the real‐time operation of urban water supply storages affected by water quality problems. We use a 3‐D, high‐fidelity simulation model to predict the main water quality dynamics and inform a real‐time controller based on Model Predictive Control. The integration of the simulation model into the control scheme is performed by a model reduction process that identifies a low‐order, dynamic emulator running 4 orders of magnitude faster. The model reduction, which relies on a semiautomatic procedural approach integrating time series clustering and variable selection algorithms, generates a compact and physically meaningful emulator that can be coupled with the controller. The framework is used to design the hourly operation of Marina Reservoir, a 3.2 Mm 3 storm‐water‐fed reservoir located in the center of Singapore, operated for drinking water supply and flood control. Because of its recent formation from a former estuary, the reservoir suffers from high salinity levels, whose behavior is modeled with Delft3D‐FLOW. Results show that our control framework reduces the minimum salinity levels by nearly 40% and cuts the average annual deficit of drinking water supply by about 2 times the active storage of the reservoir (about 4% of the total annual demand).