Control of Hybrid Microgrids Using Deep Reinforcement Learning and Digital Twin

Articles in Press

Document Type : Original Article

Authors

1 Faculty of Electrical and Computer Engineering University of Tabriz Tabriz, Iran

2 Faculty of Electrical and Computer Engineering, University of Tabriz,

Abstract

The increasing integration of renewable energy sources (RES) in hybrid microgrids has introduced new challenges in maintaining stability, reliability, and optimal performance. This paper proposes a novel control framework that combines deep reinforcement learning (DRL) with digital twin (DT) technology to address these challenges. The DRL agent is trained in a virtual DT environment, enabling rapid learning and optimization of control strategies under dynamic conditions without risking real-world operations. The proposed method is tested on a hybrid microgrid comprising photovoltaic (PV), wind, and battery storage systems. Simulation results demonstrate that the DRL-DT framework achieves a 28.5% improvement in energy efficiency compared to conventional model predictive control (MPC). Additionally, the proposed approach enhances system stability by reducing voltage fluctuations by 21.3% and achieves a 32.7% reduction in load shedding during peak demand scenarios. The training time for the DRL agent is reduced by 40% due to the efficient simulation capabilities of the DT. These results highlight the robustness and adaptability of the DRL-DT framework, making it a promising solution for next-generation hybrid microgrid management. This study provides a significant step forward in leveraging artificial intelligence and digital twins to optimize hybrid microgrid operations, ensuring sustainable and resilient energy systems.

Keywords

Main Subjects

Tabriz Journal of Electrical Engineering

Articles in Press, Corrected Proof
Available Online from 22 September 2025

Files

Share

How to cite

Statistics