Optimization of Reservoir Operation for Real-Time Flood Control with Emphasis on Forecast Uncertainty: A case study of Dez Reservoir

This study presents a novel approach for real-time flood control by optimizing operation of a dam reservoir during the flood. The real-time flood control (RTFC) with decision support systems (DSS) has recently received significant attention in flood management of reservoir dams, which results in the timely flood forecasts. However, the fundamental operational challenge of using this method is to determine optimized decisions of the reservoir management before the occurrence of flood, and optimal use of the pre-release to reduce the flood damages. In addition to this optimization challenge, since the timely forecasts are derived by employing the probabilistic methods, they would always associate with epistemic uncertainty. As a result, quantifying and considering these uncertainties in the flood control would result in reducing the risk of flood damage that can be evaluated by minimizing the expected value of the loss function represented in terms of flood damage. Timely and continuous forecasting of inflow and reservoir operation management will be implemented before the occurrence of flood until it ends. In order to address the above challenges, a novel simulation-optimization methodology, by taking into account uncertainty, is developed; and illustrated on a case study of Dez reservoir, as a highly important reservoir system in south-west of Iran, to optimize the performance of this reservoir, during the floods. Accordingly, a new water resource management has been proposed and tested. The results derived from the proposed method, indicate a significant reduction in the peak release from the reservoir, and the improved operation of the Dez reservoir in controlling the flood at real-time, which will reduce damages in the downstream area.