Optimal integrated water resources planning in watersheds with limited water resources

The literature in the area of water resources management persistently stresses the need to use an integrated water resources management approach especially in areas that experience water stress. The conventional engineering practice approach to water resources management consists primarily of collecting engineering data, evaluating a few alternatives and selecting one. In this research, optimization models for use in integrated water resources management have been developed. The first model developed, termed yearly static model, is a regional (watershed scale) water supply allocation model, which was formulated with a nonlinear objective function to maximize net benefits. The revenue from water supplies from different sources, the cost of supplying the water from these sources and the damage due to poor quality water delivered to the users were considered in the objective function. This nonlinear programming (NLP) model was applied to a regional water supply system located along the Rio Grande from Caballo dam in New Mexico to El Paso County, Texas.; In the second model, termed seasonal model', the problem was formulated as a dynamic model and applied to the regional water supply along the Rio Grande. This large-scale NLP model was too large to be solved using the previous release of GAMS/MINOS software, version 2.04, but was solvable using its latest release, version 2.25. With the same parameters considered in the objective function as the yearly static model, the solution obtained gave the "best" reservoir release policy for four seasons and a maximum net benefit that is slightly higher than that obtained using the static NLP model.; The third model developed was for the capacity expansion of water supply conveyance and treatment infrastructure over a long-term planning period. This model was formulated as a mixed integer nonlinear programming (MINLP) model that incorporates two sets of decision variables: the optimum timing of the capacity expansion of water supply conveyance systems and water treatment plants and the water allocation policy that maximizes the net benefit. A new methodology that interfaces Simulated Annealing (SA) heuristic search algorithm with the Generalized Reduced Gradient (GRG2) computer code was developed to solve this problem. The SA algorithm sets the binary decision variables used in modeling the capacity expansion.; The MINLP computer program developed is applied to El Paso County's water supply system. With three water supply regions, potentially two water sources (surface water and groundwater) for each region and a planning period of 10 years, the problem involved solving a problem size of over half a million NLP's which was formulated as a single MINLP. (Abstract shortened by UMI.)...