Deep reinforcement learning approaches for the hydro-thermal economic dispatch problem considering the uncertainties of the context

Hydro-thermal economic dispatch is a widely analyzed energy optimization problem, which seeks to make the best use of available energy resources to meet demand at minimum cost. This problem has great complexity in its solution due to the uncertainty of multiple parameters. In this paper, we view hydro-thermal economic dispatch as a multistage decision-making problem, and propose several Deep Reinforcement Learning approaches to solve it due to their abilities to handle uncertainty and sequential decisions. We test our approaches considering several hydrological scenarios, especially the cases of hydrological uncertainty due to the high dependence on hydroelectric plants, and the unpredictability of energy demand. The policy performance of our algorithms is compared with a classic deterministic method. The main advantage is that our methods can learn a robust policy to deal with different inflow and load demand scenarios, and particularly, the uncertainties of the environment such as hydrological and energy demand, something that the deterministic approach cannot do. & COPY; 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Automated summary

This study proposes Deep Reinforcement Learning approaches to solve the complex hydro-thermal economic dispatch problem, which can handle uncertainty and sequential decisions. The performance of these approaches is compared with a classic deterministic method, and the advantages of the proposed methods are highlighted.