Development of a real-time optimization model for flood control of a multipurpose reservoir

To address a problem of a multipurpose reservoir controlling flood in Taiwan, we develop a reservoir real-time operation optimization (RES-RTFC) model for flood control to identify the optimal releases during typhoon. RES-RTFC model is a combination of a series of procedures which include rainfall prediction, reservoir inflow prediction, and reservoir releases forecast. RES-RTFC model includes three sub-models: quantitative precipitation forecast average model (QPFA), streamflow prediction model (RTRL), and reservoir operation deterministic optimization model (RESOP). The objectives of RESOP sub-model include: (1) maximizing the peak flow reduction at selected downstream control-points; and (2) meeting reservoir storage at the end of the flood. The constraints incorporate the three-flood-stages operation guidelines which are suitably applied in Taiwan. The major features of RES-RTFC model are (1) determine the real-time optimal release by using the newly updated hydrological data at each time period; and (2) incorporate the reservoir guidelines of the three flood operation stages into consideration. The study area is Shihmen reservoir system. Comparing three scenarios of the historical operations, the current rales, and the RES-RTFC model running, the results show that RES-RTFC model are produce much better performances of all scenarios in terms of reducing the peak flow at the downstream control-point as well as meeting the target reservoir storage at the flood ending. Consequently, RES-RTFC model demonstrate its high practicability for estimating the optimal real-time releases.