Long-Term Generation Scheduling: A Tutorial on the Practical Aspects of the Problem Solution

Managing power production for hydro-dominated electrical energy systems in the long-term horizon is a challenging task that requires a trade-off between costs and electricity shortage risks. Because of the inflow uncertainty, this trade-off can be assessed by an operating policy that determines how much water should be used to generate power at the beginning of each time-step and how much should be stored for the future. The long-term generation scheduling (LTGS) is often solved by system operators using multistage stochastic optimization and stochastic dual dynamic programming (SDDP) approaches. Although the literature related to the LTGS problem is vast, several practical aspects used in SDDP are not demonstrated clearly, which can prevent the interest of young researchers and hinder the advances and new applications usually employed by senior professionals in industry and universities. Thus, this paper provides a tutorial on (but not limited to) (i) the application of SDDP for solving the LTGS problem, including water inflow inter-stage dependence, (ii) assessment of the impacts of hydropower function in the policy and (iii) inclusion of a risk measure to obtain a more reliable operation in unfavorable (dry) inflow scenarios. We present the main ideas behind each aspect's inclusion in the SDDP algorithm, using simple (but comprehensive) numerical examples. The overview is intended to understand these main characteristics, presenting them as powerful tools to investigate the LTGS problem in detail, searching for a good or even (near) optimal operating policy.