Continuous re-optimization of reservoir policies as an adaptation to climate change
Abstract
Adapting reservoir control policies is a promising approach to managing uncertain climate change within the limits of existing infrastructure. However, it has not yet been investigated how control policies can be dynamically adapted as new observations of nonstationary flood and drought risk become available. Here we develop a continuous re-optimization approach in which control policies are adapted with a fixed frequency (f) using a recent window (w) of observations. The approach is demonstrated for the multi-reservoir system in the Sacramento-San Joaquin River basin, California. Control policies are structured as parameterized rules, and a climate model ensemble is used for forcing scenarios. For each reservoir, we test a range of frequencies from 1-20 years and window sizes from 5-50 years to find the best combination for re-optimization by balancing the tradeoff between adapting quickly versus correctly. Out-of-sample performance is evaluated during the intervals between optimizations using a daily timestep simulation model. Through this experiment, we find how the reservoir operating curve parameters change over time, and how this modifies typical storage trajectories in wet and dry scenarios. We also compare the results of re-optimization with the simulation of the calibrated historical policy, and with a perfect foresight optimization. Reliability of water supply for re-optimization scenario is found be higher than that of simulated historical policy. The proposed methodology enables reservoir operators to develop a strategy for regularly adapting control policies based on nonstationary hydrology by choosing the frequency of optimization and maximizing the performance of the system.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H35F..04S