A Framework to Provide Optimal Management Strategies for California’s Reservoirs in Achieving Sustainable Water Supply and High Hydropower Productivity
- Author(s): YANG, TIANTIAN
- Advisor(s): Sorooshian, Soroosh
- Gao, Xiaogang
- et al.
With the increasing demands on freshwater water and clean energy due to population growth and impacts of climate change, the stresses on natural resources are increasing worldwide. Therefore, efficient operation of reservoir systems with the intention of optimizing sustainable water supply and hydropower production is crucially needed by policy and decision makers, and water users. In this dissertation, a framework, including analysis of reservoir controlled outflows, optimization algorithm development, and realistic reservoir modelling, is presented and demonstrated in Chapter 2, 3 and 4, respectively.
In Chapter 2, a Classification And Regression Tree (CART) algorithm with an enhanced cross-validation scheme is applied to simulate the human controlled outflows in 9 major reservoirs in California. The proposed approach is capable of incorporating multiple types of information into decision making and mathematically quantifying how releases are related to many decision variables. A verification study has been carried out in 9 major reservoirs in California. Without any prior information, the model is able to identify that the historical operation in Oroville Lake, Shasta Lake and Trinity Lake are highly dependent on policy and regulation, while the reservoirs with low elevations are sensitive to reservoir inflows. The approaches developed in this chapter serves as the analytical tool to help understand reservoir operation.
In Chapter 3, an enhanced multi-objective global optimization technique is developed in order to better address multiple conflicting interests from decision makers when water and energy related objectives are jointly considered in reservoir operation. A comparison study has been conducted comparing the enhanced algorithm with multiple cutting-edge multi-objective heuristic search algorithms on various test functions. Results show the enhanced algorithm has superior performance regarding diversity and convergence measures over the other algorithms.
Last, a newly developed cascade reservoir optimization model for the Oroville-Thermalito Complex (OTC) in northern California is presented in Chapter 4. Multiple alternative operation strategies that maximize sustainable water supply and hydropower production are derived and recommended for the OTC’s operation under various dry/wet conditions. The suggested optimal operation alternative will be intuitive for reservoir operators to further adjust and improve current reservoir operation strategy and planning.