Combined Application of Dynamic Programming and Network Flows Optimization for River Operation Simulation

Project ID: 924
Principal Investigator: Roger Larson
Research Topic: Water Operation Models and Decision Support Systems
Funded Fiscal Years: 2004
Keywords: None

Research Question

What approach can be developed that combines the strengths of time-horizon optimization, available in 'dynamic programming' models, and the prioritized distribution of flows, available in 'linear network flows' models?

"Time-horizon optimization" refers to a model's ability to look ahead and make current time-step decisions based on future decisions and outcomes. "Prioritized distribution," in the case of river and reservoir simulation, refers to the ability of the model to manage flow distribution based on the Doctrine of Prior Appropriation.

Need and Benefit

'Dynamic programming' optimization offers an effective means to derive reservoir rule curves and release schedules over a time horizon because it has the ability to select the best current decision based on future decisions and events. The result of a "dynamic programming" approach is the optimal solution over time. However, "dynamic programming" has limitations in scaling to systems with a large number of reservoirs and demands.

In contrast, "linear network flows" optimization efficiently distributes flows based on priorities or water rights dates for large-scale river and reservoir systems, but it's knowledge of the system is limited to the current time step. Although all river systems models produce solutions over space, the result of a "linear network flows" approach is the optimal solution over space.

In combining the strengths of each of these optimization techniques, the authors propose to replace the trial and error approach of developing model rule curves and operating criteria with a single, less time-consuming, more accurate approach. The solution will be optimal or near-optimal over both time and space. Once the approach is developed, it can be applied generically to all modeled systems. Greater accuracy and less model development time leads to more time available for resource analyses and a faster response time to study requests.

Contributing Partners

Contact the Principal Investigator for information about partners.

Research Products

Please contact about research products related to this project.

Return to Research Projects

Last Updated: 4/4/17