COMPUTATIONALLY EFFICIENT HYDROPOWER OPERATIONS OPTIMIZATION FOR LARGE CASCADED HYDROPOWER SYSTEMS REFLECTING MARKET POWER, FISH CONSTRAINTS, MULTI-TURBINE POWERHOUSES, AND RENEWABLE RESOURCE INTEGRATION

Abstract

Hydropower generation, though a centuries-old technology, is gaining new relevance as a way to integrate renewable energy sources to the power grid. This dissertation describes the development of efficient models for optimizing hydropower operations to address the competing priorities of renewable generation integration, economically efficient hydropower generation, and environmental stewardship, while still being able to maximize the value of wind and hydropower generation. The models are demonstrated using the 10-reservoir Federal Columbia River Power System in the Pacific Northwest in the U.S.A. Our computationally efficient nonlinear optimization model for the 10-reservoir system employs variable time step lengths and precomputed powerhouse functions and operating rules. Having shorter 8-hour time steps in the first few days of the model then transitioning to a coarser 24-hour time step for flow routing in the later stages results in optimization runtimes being decreased to 1/3rd to 1/6th of the time it takes to run the optimization with all 8-hour time steps. Our powerhouse functions reduced the many dispatch and loading decisions for multiple turbines at a hydropower project into a powerhouse generation as a function of total flow. The nonlinear optimization model incorporates forecasted inflow, hydropower plant operation, contracted energy loads, and the hydropower utility’s interaction with wholesale energy markets. When applicable the model also includes special seasonal constraints for fish addressing specified turbine operations and upper and lower bounds on spils. The opportunity costs of meeting these environmental constraints can be estimated. Additionally, we consider the market power of a very large hydropower producer in a regional market. For an entity with market power, maximizing avoided cost will result in prices that are very similar across periods, which is the economically efficient solution. In contrast, maximizing revenue will result in prices that are not balanced across periods, which may result in monopolistic behavior. To address renewable integration, our stochastic dynamic programming and nonlinear programming model builds upon the aforementioned framework with a time decomposition approach to optimize the hourly operations of a subset of the 10-reservoir hydropower system under wind generation uncertainty. This model also includes the effect that wind generation has on market prices, in addition to the hydro utility market power. We showed how introducing increasing levels of wind generation uncertainty causes the model to hedge by decreasing its commitment to the wholesale electricity market. The model estimates the opportunity costs of providing hour-by-hour balancing of the wind generation to a wind power generation owner.