Medium-Term Hydropower Scheduling: In a Multi-Market Setting

Regulated hydropower is a flexible resource that is well suited for the provision of ancillary services. Given the opportunity to participate in different markets, the difficult task is to find the optimal allocation between them. The scope of this thesis is to investigate the multi-market problem and provide models for the optimal utilization of regulated hydropower in the Medium-Term Hydropower Scheduling (MTHS) setting. This is conducted from the view of a hydropower producer, participating in the day-ahead electricity market and in reserve capacity markets for providing rotating reserves. A major part of Norway’s electricity generation comes from hydropower. Since the annual inflow displays considerable fluctuation, it is imperative that hydropower reservoirs are efficiently managed. The stored water should be exerted when it is required, i.e. when the market prices are high compared to the value of storing it for later use, at the same as spillage should be avoided andmarket revenues maximized. With better grid connection to the rest of Europe and tighter market coupling, flexible hydropower producers have an edge in providing their flexible resources to larger market shares. The tighter market coupling also ensures a higher security of supply in the Nordics and the ability to absorb more renewable energy at times when there is, alternately, a great deal of wind or sun in continental Europe. The objective of this thesis is thus to provide methods for decision support for hydropower producers in a changing power market. The initial work in this thesis investigates how effectively a current Medium-Term Hydropower Scheduling (MTHS) model, based on Stochastic Dual Dynamic Programming (SDDP), performs in a multi-market setting. The SDDP algorithm is a state-of-the-art method for solving multistage stochastic programming problems with extensive adoption for Hydropower Scheduling (HS) problems. The work found that the SDDP model overestimated the hydropower system’s ability to provide reserve capacity. For the given case studies it was around 30%, illustrating the importance of detailed modeling when considering reserve capacity. To undertake the issue with modeling details the newly proposed Stochastic Dual Dynamic integer Programming (SDDiP) method was applied to the MTHS problem. It was shown that the method provided convergence of a nonconvex MTHS problem, but with a substantial computational burden. A new type of cut in the SDDiP framework, called a strengthened Benders cut, showed beneficial properties in terms of an improved optimality gap and manageable computation time. These results were further sustained by another study that included more detailed modeling of the hydropower system with a recently proposed method to include uncertainty in objective term coefficients for Dynamic Programming (DP) problems. The study showed that an approach based on Benders cuts was inferior when the problem formulation was so complex. Furthermore, a study of a constructed power system with increasing shares of wind power was investigated, with an emphasis on the provision of both upwards and downwards reserve capacity from hydropower and wind power. The study was conducted with a SDDP model showing that wind power could effectively provide downwards reserve capacity, and in certain cases when the wind penetration was very high some upwards reserve could also be provided. Based on the possibility of including nonconvexities in the MTHS problem with the SDDiP method, a study on the modeling of the generation function from a hydropower station was also conducted. It was found that the convex relaxation of this function leads to an overestimation of what is physically possible. This is particularly true when environmental constraints and a reserve capacity market was included. In short, this thesis investigated both current and recently proposed algorithms used for solving the MTHS problem. Realistic case studies were applied with the aim of presenting state-of-the-art algorithms that are applicable in operational use. This is especially evident for an improved type of Benders cuts called Strengthened Benders cuts, used by the SDDiP method