Use of stochastic dynamic programming for reservoir management

Publisher Website
Google Scholar
Abstract
Hydropower production plays an important role in the operation of electric energy production‐distribution systems. Since most of the economically feasible hydroelectric sites have already been developed, it is necessary to examine and develop practical, real‐time operational models which can be used to increase the output from existing hydropower plants. Three main issues are addressed by this research: the potential of increasing the output from existing hydropower plants, the alleviation of dimensionality problems for multistate dynamic programming, and the use of probabilistic forecast in the decision‐making process. An optimization model is developed which can be used as an analytical tool in the decision‐making process for reservoir operation. The model takes into consideration the uncertainty of forecast at the time a policy must be established. The uncertainty is expressed in terms of the second moments of the forecast probability distributions. There is no limitation on the type of distribution, and it is assumed that forecast is made by a conceptual type of watershed model. The proposed methodology is applicable to constrained stochastic systems with quadratic objective functions and linear dynamics. It uses the decomposition principle of dynamic programming without discretizing the state or control variable and therefore the method can be used for large‐scale systems. It is an iterative procedure which requires an initially feasible solution and solves a series of quadratic programming problems at each iteration. The applicability of the research is demonstrated through case studies.

This publication has 14 references indexed in Scilit: