Optimal Stochastic Operation of Salt River Project, Arizona

Abstract

The application of extended linear quadratic Gaussian (ELQG) control, a nonlinear stochastic optimization method, to the Salt River Project reservoir system is described. The optimization problem was posed as the maximization of total hydropower avoided cost, subject to constraints that represented system mass balance and physical or operating limitations on releases. The six surface reservoirs, ground‐water storage, and planned cyclic ground‐water storage were modeled as six equivalent reservoirs. The ELQG control algorithm was applied to the historic reservoir inflow for the 51‐yr period 1931–1981 using a 12‐month sliding window. Operating policies for each of the 600 months in the test period were developed and compared with the present operating policy for two reservoir configurations: the existing configuration, and a reduced storage configuration that mimicked temporary system constraints during the construction period for the planned alteration of Roosevelt Dam. The ELQG control results showed that significant increases in hydropower avoided costs could be achieved while maintaining water supply reliability at current levels or higher.