Decentralized Reinforcement Learning Approach for Microgrid Energy Management in Stochastic Environment

Microgrids are considered to be smart power grids that can integrate Distributed Energy Resources (DERs) in the main grid cleanly and reliably. Due to the random and unpredictable nature of Renewable Energy Sources (RESs) and electricity demand, designing a control system for microgrid energy manage...

Full description

Saved in:
Bibliographic Details
Published in:International transactions on electrical energy systems 2023, Vol.2023, p.1-15
Main Authors: Darshi, Razieh, Shamaghdari, Saeed, Jalali, Aliakbar, Arasteh, Hamidreza
Format: Article
Language:English
Subjects:
Algorithms
Communication
Control systems
Customers
Decision making
Deep learning
Design
Distributed generation
Electric power demand
Electric vehicles
Electricity
Energy management
Energy policy
Energy resources
Energy storage
Games
Multiagent systems
Neural networks
Profits
Renewable energy sources
Teaching methods
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Microgrids are considered to be smart power grids that can integrate Distributed Energy Resources (DERs) in the main grid cleanly and reliably. Due to the random and unpredictable nature of Renewable Energy Sources (RESs) and electricity demand, designing a control system for microgrid energy management is a complex task. In addition, the policies of microgrid agents are changing over time to improve their expected profits. Therefore, the problem is stochastic and the policies of the agents are not stationary and deterministic. This paper proposes a fully decentralized multiagent Energy Management System (EMS) for microgrids using the reinforcement learning and stochastic game. The microgrid agents, comprising customers, and DERs are considered as intelligent and autonomous decision makers. The proposed method solves a distributed optimization problem for each self-interested decision maker. Interactions between the decision makers and the environment during the learning phase lead the system to converge to the optimal equilibrium point in which the benefits of all the agents are maximized. Simulation studies using a real dataset demonstrate the effectiveness of the proposed method for the hourly energy management of microgrids.
ISSN:2050-7038
2050-7038
DOI:10.1155/2023/1190103