Integrating Machine Learning Techniques into the Decision-making Process for Hydro Scheduling

Abstract

During the past half-century, numerous optimization models have been developed to help hydropower producers to determine the optimal power generation schedules. Nevertheless, the producers must manually set up the executive commands before running the optimization models. Limited by human analytic competence, the producers usually use the default setting. The value of the optimization tools could be further carried forward if the commands are dynamically determined according to the specific operating and market conditions. In this paper, we propose a framework and methodologies to facilitate the decision-making process for hydropower producers by realizing the automatic setup of executive commands. This automation is achieved by integrating machine learning (ML) techniques with a comprehensive understanding of the hydro systems and the hydro scheduling tools. It is demonstrated that nonphysical spills from reservoirs can be 100% avoided using the command setting predicted by ML compared to the result obtained by the default setting. The calculation time can reduce by 45% compared to the robust setting.