Reinforcement Learning Solutions for Microgrid Control and Management: A Survey

Author(s)

Barbalho, Pedro I. N. ; Moraes, Anderson L. ; Lacerda, Vinicius A. ; Barra, Pedro H. A. ; Fernandes, Ricardo A. S. ; Coury, Denis V.

• Document type: Journal article
• Source: IEEE ACCESS; v. 13, p. 18-pg., 2025-01-01.

Abstract

A microgrid (MG) is part of a distribution system that comprises loads and distributed energy resources, capable of operating either connected to or islanded from the primary grid. Having an appropriate design, MG controllers improve energy efficiency, playing a vital role in the modern distribution system. Thus, MG management and control has become a broad area of research due to its complex operation. Reinforcement learning (RL) offers adaptive solutions for handling MG complex dynamics and nonlinearity. It is an alternative to traditional algorithms and control methods in tasks, such as load frequency control, resource allocation, and energy management. Due to the relevance of the topic, this survey examined the role of RL in MG control and management, offering a comprehensive update on previous reviews, categorising articles by RL type, control objectives, and MG operational modes. Additionally, hardware implementations and performance assessments across RL-based solutions were evaluated. The present survey identified key research trends and gaps, contributing to understanding the role of RL in MG management and control and guiding future solutions in the field. (AU)

• FAPESP's process: 23/00182-3 - Harmonic source location and identification: an approach embedded in low-cost meter
• Grantee: Ricardo Augusto Souza Fernandes
• Support Opportunities: Regular Research Grants
• FAPESP's process: 21/04872-9 - Evaluating technical impacts on the expansion of distributed generation for smart grids
• Grantee: Denis Vinicius Coury
• Support Opportunities: Regular Research Grants