A Novel Physically-Based Framework for the Intelligent Control of River Flooding

Presented at The Oregon Water Conference, May 24-25, 2011, Corvallis, OR.River flooding is a recurrent threat and its control and management continues to be a challenge. It has been recognized that effective flooding control requires a real-time strategy that combines optimization with a physically-based simulation model. Current real-time frameworks that combine simulation and optimization have two main drawbacks. The first drawback is that they attain the best operation strategy based on short-time forecasting only (few hours – few days). This may lead to a wrong operation strategy that may result in flooding or to an unnecessary water release from the reservoirs which would be in conflict with non-real time objectives of the system, such as those of maximizing water storage for irrigation and hydropower production. The second drawback is that they do not account for system flow dynamics. These frameworks instead simply perform mass balance analyses in the reservoirs and assume that the water levels in the reservoirs are horizontal. This is a strong limitation given that a flooding event is highly dynamic and may start from anywhere in the river system. It may start from upstream (e.g., large inflows), from downstream (i.e., high water levels at downstream) or laterally from the connecting reaches (e.g., water levels at river junctions are near the reach banks). Accounting for system flow dynamics is also important because the flow conveyance from one reservoir to another is not instantaneous but depends on the capacity of the connecting reaches, the capacity of the associated gates and outlet structures and the dynamic hydraulic gradients. We present a novel simulation-optimization real-time framework that (1) accounts for system flow dynamics, (2) maximizes the benefits of non-real time objectives of the regulated river system at all times, except during a period, determined automatically trough sampling for a long forecasting, in which the objective of the system will switch to minimize flooding and (3) allows controlled flooding only after the capacity of the entire river system has been exceeded. This controlled flooding is based on hierarchy of risk areas to losses associated with flooding. This means that river reaches (or their areas of influence) that are less prone to losses will be assigned higher preferences for locations of flooding. Once the sampling determines that there is no danger for flooding, the proposed framework automatically switches to maximizing the non-real time objectives of the system. This sampling accounts for unsteady boundary conditions and for system flow dynamics, in particular for computing the conveyance capacity of the system. We demonstrate the proof of concept of this new framework using a hypothetical river system