Multi-objective reservoir operation during flood season considering spillway optimization

Flood control and hydropower generation are two main functions of Three Gorges Reservoir (TGR) in China. In this study, a multi-objective operation model for TGR considering these two functions was developed. Since the optimal results of reservoir operation are mostly in the form of gross outflow which is hardly used to directly guide reservoir operation, the optimization of spillways operation was taken into account. For observed historical flood hydrographs and design flood hydrographs, the progressive optimality algorithm (POA) was employed to determine the optimal operation of spillways. For the real-time reservoir operation, a smooth support vector machine (SSVM) model was applied to abstract the optimal operation rules which consider the order and the number of spillways put into use. Results demonstrate that the use of different spillways has a significant impact on reservoir operation. Therefore, it is necessary to consider the order and number of spillways that should be used. Instead of optimizing outflow, direct optimization of the order and number of spillways can yield most reasonable results. The SSVM model simulates the relationship among inflow, water level and outflow satisfactorily and can be used for real-time or short term reservoir operation. Application of the SSVM model can also reduce flood risk and increase hydropower generation during the flood season. 2017 Elsevier B.V. All rights reserved.