Modeling the Inference between Upstream Inflow Hydrographs and Downstream Flooded Areas in a Reservoir Driven System

Management of a river imposes a task for several institutional levels in every part of the world, especially when reservoirs are the main flow drivers. The research presented herein shows how management of the flow in a reservoir driven system can be modeled by using an inference simulation model between two sets of variables, as an alternative to an approach based on an optimization model. The first selected variables are the explanatory ones, which refers to the relation between the operation of reservoirs and the input hydrographs to each of the considered reservoirs. The second set of variables are descriptive and refers to the resulting flooding hydrograph at a hydrological station located downstream of the reservoir network. The Yellow River, in China, is chosen to demonstrate the concept, however the methodology can be applied in practice for any reservoir driven system. In the mid stream of the Yellow River, a system of four reservoirs was built to manage flooding, using daily information and taking daily decisions. The proposed method uses a simplified simulation model of the actual reservoir operation to determine a multiple linear regression model between the set of explanatory and descriptive variables. The set of explanatory variables is very large and the demonstration is done on a selected subset of variables. Each selection of an explanatory variable is based on a correlation analysis with respect to the original set. Analysis of the model results shows that the reduction of the number of variables does not decrease the model fitness and robustness.