| A Multi-Reservoir Testbed in Borg-RiverWare to Aid Multi-Objective Simulation-Based Optimization |
|
Joseph Kasprzyk—CU Boulder; Edith Zagona, David Neumann, & Patrick Lynn—CADSWES |
 |
ABSTRACT |
|
|
Multi-reservoir river systems provide benefits to diverse water uses, accrued to different stakeholders that have interests at different points in the reservoir network. Simulation-based optimization using Multiobjective Evolutionary Algorithms (MOEAs) generate new alternative management plans for such problems, representing tradeoffs among multiple competing objectives. MOEAs' performance is tested using test problems, either sets of mathematical equations that are known to be difficult to solve, or real-world inspired problems (e.g. the Lake Problem for environmental management). Although many test problems have been presented in previous work, there is a need to create a test problem that links more directly to real-world reservoir management. For example, simulations in water resources planning generate a large amount of timeseries data for multiple variables across the nodes of the network. Quantitative metrics such as reliability, resilience, and vulnerability can be calculated to help summarize alternatives' performance, but it is challenging to find the proper set of performance metrics given multiple possible performance targets at the different nodes within the system. This presentation will introduce a realistic multi-reservoir system that was modeled using the Borg-RiverWare software, which integrates the Borg MOEA with RiverWare, a river basin modeling software platform. By contributing the testbed in RiverWare, we demonstrate how the interactive visualizations and policy language functionality can aid the process of formulating a multi-objective optimization problem that best meets diverse needs of water managers. |
|
