Modeling the California State Water Project

Abstract

Two disadvantages of optimization models in water resources planning and management are the simplifications often required to construct the model and user reluctance to rely on optimization models alone. Both of these disadvantages may be overcome by combining a simulation model with an optimization model. In this study, a popular multipurpose, multireservoir system simulation model, HEC‐3, developed by the Hydrologic Engineering Center (HEC) of the Corps of Engineers, is combined with a dynamic programming model. One problem in using discrete differential dynamic programming (DDDP) is determining the initial trajectory. In this project, conventional dynamic programming was used to estimate the initial trajectory, and then a DDDP model was used. Consequently, the optimization model combines conventional dynamic programming and discrete differential dynamic programming to solve the complex problem of operating the California State Water Project's (SWP) Lake Oroville and San Luis Reservoir, given the constraints of the Federal Central Valley Project (CVP) and delta flow requirements. Preliminary results suggest that SWP revenue may be almost doubled by adopting the suggested optimal operation.