Optimal control of multiple reservoir river systems using nature based algorithms

The objective of this research is the development of a discrete time optimal control methodology and associated model that indirectly minimizes the negative impacts on aquatic flora and fauna caused by unnatural water level fluctuations that may occur in multiple reservoir river systems, while continuing to meet the intended purposes of the reservoirs. The optimal control approach employed integrates the National Weather Service's unsteady, one-dimensional, comprehensive hydraulic simulation model, FLDWAV, with three alternative nature based optimization/search tools, namely genetic algorithm (GA), a simulated annealing (SA) algorithm, and an artificial life algorithm (ALA). In addition to hydraulic constraints that are handled by the simulation model, five operational constraints are considered, including bound constraints on water levels at desired cross sections; storage levels in reservoirs; releases from dams; gate openings; and minimum pool levels behind locks and dams. Constraints on decision variables, dam releases and gate openings, are handled directly by the optimization algorithm, while those on the state variables, including storage levels, water levels, and minimum pool level requirements, are handled through a penalty function method. Furthermore, a statistical preprocessing tool is integrated for automatic evaluation of lower and upper water level bounds. The resulting model, GA/SA/ALA-FLDWAV, has the capability to optimize the operation of multiple reservoir river systems under three types of reservoir operation modes; namely, (i) head independent releases, (ii) channel/gate controlled flow that corresponds to flow through Wicket type dams, and (iii) head dependent releases corresponding to flows through Tainter gates. To evaluate the models capacity (develop coordinated, optimal reservoir operation policies), it has been applied to three systems, a hypothetical three-dam, two-river system, the Peoria-LaGrange reach of the Illinois River, and a longer, 425 km section of the Illinois River for which the simulation model was calibrated and verified. Results of application of the model on these systems suggest the robustness of the model in identifying the global or near global solution. The model has a potential to assist reservoir operation managers in developing coordinated reservoir operation policies that meet environmental requirements, while satisfying the intended purpose(s) of the reservoirs.