Optimal operation of multiple reservoirs in hydropower-irrigation systems : a stochastic dual dynamic programming approach

The increasing global water demand due to population growth and higher living standards exerts a significant stress to the limited global freshwater resources. Meeting the future global water demand therefore implies that the operational water allocation policies are efficient and effective. Computational tools relying on optimization algorithms can be used to derive efficient allocation policies. Yet, due to computational constraints, existing tools have only been applied to the analysis and design of small-scale water resources systems. As integrated water resources management requires that allocation policies be determined at the river basin scale, traditional optimization models have limited applicability due to the high dimensionality of basin-wide allocation problems. The overall objective of the thesis is to contribute to the improvement of the operational efficiency and effectiveness of existing and planned water resources systems and particularly the reservoirs operation. The specific objective is to adapt, to test and to assess the usefulness and applicability of the Stochastic Dual Dynamic Programming (SDDP) algorithm for determining economically efficient water allocation policies in stochastic multipurpose multireservoir systems. SDDP has been designed to circumvent the curse dimensionality problem associated with traditional optimization methods. More specifically, SDDP is used to analyze the hydrologic and economic consequences associated with various development and management scenarios in large-scale river basin systems in which hydropower generation and irrigated agriculture are the main uses. The algorithm has been successfully implemented and evaluated for analyzing water allocation in two case studies: the Eastern Nile and Euphrates river basins.