Back to search

Georgia Institute of Technology

State form dynamic programming: a new optimization method for large reservoir systems

Abstract

dc:description.abstract

Numerous optimization methods have been proposed for reservoir management in the last several decades. Although these methods claim superior solutions compared to simulation-based management procedures, their adoption in practice has been hindered due to their simplified system representations and excessive computational requirements. This thesis introduces and evaluates a new optimization approach (State Form Dynamic Programming, SFDP) designed to address these challenges and facilitate efficient management of large, real-world reservoir systems. SFDP avoids full-space discretization by decomposing the optimization process into two iterative steps aiming to (1) optimize total system storage sequence and (2) distribute the system storage optimally among the individual reservoirs at any point in time. It is shown that this iterative strategy identifies optimal solutions while drastically reducing the required computational effort. To validate its performance, SFDP is benchmarked against well-known optimization methods including Dynamic Programming (DP), Discrete Differential Dynamic Programming (DDDP), and Genetic Algorithms (GA) in systems of increasing size from 2 to 10 reservoirs. These systems are portions of the ACF and ACT Basins in the southeast U.S. The system performance is assessed across multiple realistic metrics reflecting water, energy, and environmental management objectives. SFDP is shown to outperform other optimization methods, especially for systems of more than 3-4 reservoirs, where it quickly becomes the only viable optimization option. The SFDP practical potential is further demonstrated through comparison of its policies with those encoded in ResSim, an industry simulation tool standard developed by the U.S. Army Corps of Engineers. SFDP consistently identifies: (1) superior trade-offs between dependable capacity and prime energy generation during wet, normal, and dry periods, (2) substantially increases primary energy production and economic value, (3) maintains higher mean and minimum system storage while fulfilling other requirements, (4) achieves better drought management, and (5) protects aquatic systems with higher reliability. Lastly, the SFDP applicability under uncertain hydrologic conditions is demonstrated and evaluated through a retrospective assessment of a coupled inflow forecasting - SFDP model. The inflow forecasting model utilizes a Markov Chain framework and the assessment is carried out for the ACF Basin. Notwithstanding forecast uncertainty, SFDP meets prime energy targets and sustains greater system storage with higher reliability than ResSim, confirming its high potential for use in the management of real‑world reservoir systems.

Degree

thesis:*
Level thesis:degree_level
Doctoral
Department dc:contributor.department
Civil and Environmental Engineering
Grantor dc:publisher
Georgia Institute of Technology
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Chen, Lida
Advisor dc:contributor.advisor
  • Georgakakos, Aristidis P.
Committee members dc:contributor.committeemember
  • Luo, Jian
  • Wang, Jingfeng
  • El Sharif, Husayn
  • Zeng, Wei

Subjects

dc:subject × 7

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/82047

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Chen, Lida. State form dynamic programming: a new optimization method for large reservoir systems. Doctoral thesis, Georgia Institute of Technology, 2025. https://hdl.handle.net/1853/82047