Adaptive Multi-Objective Energy Management in Renewable-Integrated Microgrids Using Deep Q-Network Learning

作者：Anshad A S, Mohammed Saleh Al Ansari, Deepanshi, Sumathy Muniamuthu, Sunny Vig, Sachin Kumar · 发表于：2026 1st International Conference on AI, Data Science, Cyber Security and Smart Manufacturing for Sustainable Development (ICADCS) · 年份：2026 · DOI：10.1109/icadcs70036.2026.11583036

Energy management in renewable-integrated smart microgrids is difficult because of the uncertainty in nature of renewable sources as well as in dynamic load variations. The growing adoption of renewable energy sources in the contemporary energy systems presents operational ambiguities and energy management problems in microgrids. The usual rule-based and optimization-based energy management techniques usually cannot adjust to the dynamic variations in loads and stochastic renewable generation. The proposed paper is a Deep Reinforcement Learning (DRL) dependent Energy Management Strategy (EMS) of a renewable-integrated smart micro grid, which includes photovoltaic (PV) generation, Battery Energy Storage System (BESS), utility grid as well as dynamic load. The model proposed is written and tested on MATLAB with the aid of the Reinforcement Learning Toolbox. A Deep Q-Network (DQN) agent is an agent that is aimed at reducing grid reliance and costs of operation without violating the battery State of Charge (SOC) limits. The two controllers compared are the proposed DRL-based EMS and a traditional rule-based controller in terms of their performance. Simulation results demonstrate that the proposed DRL-based EMS achieves approximately $\mathbf{1 8} - \mathbf{2 2} {\%}$ reduction in operational cost, reduces grid dependency by nearly $15-20 \%$, and improves renewable energy utilization by as much as $\mathbf{2 5 \%}$ compared to conventional rule-based methods. Additionally, the sy...