AI-Powered Microgrid Networks: Multi-Agent Deep Reinforcement Learning for Optimized Energy Trading in Interconnected Systems

• Published in: Arabian journal for science and engineering (2011) Vol. 50; no. 8; pp. 6157 - 6179
• Main Authors: Alferidi, Ahmad, Alsolami, Mohammed, Lami, Badr, Slama, Sami Ben
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2025; Springer Nature B.V
• Subjects: Artificial intelligence; Deep learning; Distributed generation; Electric power demand; Electric vehicles; Energy consumption; Energy costs; Energy management; Energy requirements; Engineering; Humanities and Social Sciences; Machine learning; Multiagent systems; multidisciplinary; Peer to peer computing; Real time; Renewable energy sources; Research Article-Electrical Engineering; Residential energy; Science; Time of use electricity pricing; Deep reinforcement learning; Electric vehicles; Distributed generation; Real-time pricing; Intelligent microgrids; Peer-to-peer energy trading
• ISSN: 2193-567X; 1319-8025; 2191-4281
• DOI: 10.1007/s13369-024-09754-4

Intelligent smart microgrids have been identified as a subject of significant research interest, given their potential to optimize energy consumption in residential contexts. The growing utilization of intelligent appliances and the integration of renewable energy sources, including distributed generation (DG) and electric vehicles (EVs), have increased energy demand. This paper presents an artificial intelligence (AI) system that employs deep reinforcement learning to facilitate efficient device scheduling and peer-to-peer (P2P) energy trading within microgrids. The system accommodates users with varying access levels to distributed generation (DG), battery storage, and electric vehicles (EVs). The real-time pricing and demand response mechanisms enable the system to adapt to fluctuating energy requirements. In contrast, surplus energy is shared through a peer-to-peer network, reducing grid dependency. The approach was validated using an experimental database from Saudi Arabia, demonstrating a notable reduction in electricity costs for participants.