Real-time operation of reservoir systems, information uncertainty, system representation and computational intractability
Teegavarapu, Ramesh S. V.
MetadataShow full item record
Modeling real-time reservoir operations and developing optimal rules are formidable tasks considering a number of issues that need to be addressed within optimization and simulation models. The issues range from uncertain system inputs to implementation of operating rules in real-time. This dissertation addresses some of these issues that are relevant at different stages of real-time reservoir operation process. These issues are: (i) information uncertainty; (ii) system representation; and (iii) computational intractability. Realtime operation models are developed in the present research for single and multiple reservoir systems while addressing these issues in that order. Uncertainty generally associated with system variables in a variety of forms is a main hurdle in developing a proaches for optimizing reservoir operations. Explicit and implicit stochastic approaches based on traditional probability theory concepts cannot always handle all the uncertain elements of reservoir operation. Approaches to handle imprecise information are required as much as methodologies to address the issue of lack of information. The former issue described as information uncertainty in this thesis is addressed using fuzzy set theory. Mathematical programming models are developed under fuzzy environment to handle imprecise and uncertain components of reservoir operation problem dominated by an economic objective. The concept of 'compromise operating polices' is proposed and its utility is proved. Representation of physical system in mathematical programming formulations affects the extent to which the physics of the problem is captured and nature of the solutions that can be obtained. Tradeoffs between exhaustive representation and optimal solutions can be identified. Operation of a multiple reservoir system is considered to develop formulations of varying degree of system representation. A Mixed Integer Non-Linear Programming (MINLP) Model with binary variables is developed to a specific case of coupled hydropower reservoirs. The model is considered to be innovative in handling the issue of hydraulic coupling. In addition to this, a new model is also proposed for the same problem based on a spatial decomposition approach. These formulations can be superior to the already existing approaches in the literature. Classical optimization techniques fail to provide solutions to mathematical programming formulations whenever an exhaustive representation of the physical systems is considered. This is due to large number of variables often part of formulations at finer time scales, special conditions and variables. This problem is referred to as computational intractability. To handle this issue, an optimization model based on a stochastic search technique (' Simulated Annealing') is proposed. The approach with few conceptual improvements is applied to multiple reservoir operation problems plagued by dimensionality and computational intractability issues. Application to standard benchmark and real-life problems confirms the immense potential this approach. holds for intractable reservoir operation problems. Simulation models and a support system that aid the decision making process of reservoir operators in realtime are also developed as a part of research work. The simulation models are developed using an Object-Oriented simulation environment that is based on the principles of System Dynamics. A Decision Support System (DSS) encompassing all the simulation and optimization models developed in the present research is designed and implemented. Some features of these approaches are considered innovative and practical. These approaches can help address the issues of actual implementation of operating rules and bridge the gap between theory and practice.