Optimal Real-Time Control of Stochastic, Multipurpose Multireservoir Systems
AIR FORCE RESEARCH LAB BOLLING AFB DC
Pagination or Media Count:
This thesis presents new systems analysis methods that are appropriate for complex, nonlinear systems that are driven by uncertain inputs. These methods extend the ability of discrete dynamic programming DDP to system models that include six or more state variables and a similar number of stochastic variables. This is accomplished by interpolation and quadrature methods that have high-order accuracy and that provide significant computational savings over traditional DDP interpolation and quadrature methods. These new methods significantly improve our ability to apply DDP to large-scale systems. Using these methods, DDP can solve a variety of systems analysis problems without resorting to the simplifying assumptions required by other stochastic optimization methods. This is demonstrated in the application of DDP to problems with as many as seven state variables. Of particular interest, this thesis applied DDP to the practical problem of conjunctively managing groundwater and surface water. Moreover, the applications also demonstrate that DDP can be a powerfill planning tool, such as when evaluating a range of capacity expansion alternatives.
- Operations Research
- Computer Programming and Software