| Abstract: |
The growth in the amount of distributed energy resources and the rise of modern complex power grids has led to the creation of smart Microgrid Energy Management Systems (MEMS). In this invited review paper, we perform a met study of AI and automation methods used in energy management for microgrids, which abstracts from 30+ years (1992 - 2024) of studies. This review presents a systematic review of the use of ML algorithms, deep reinforcement learning, fuzzy logic controllers (FLCs), multi-agent systems and hybrid artificial intelligence approaches for energy scheduling, demand response and grid stability in islanded microgrids as well as in grid-connected microgrids. The review indicates a shift from traditional rule-based controllers to data-driven intelligent frameworks only that are able to manage the stochastic characteristics of renewable energy resources, time-varying load demands and dynamic market pricing. A quantitative synthesis of the published results shows that mems with custom artificial intelligence (AI) trained methods consistently achieve 15-35% lower energy costs and 20-40% higher operational efficiency than traditional approaches. Cybersecurity resilience, real-time scalability, and standardized benchmarking have large critical gaps. In the final part of the paper, a forward-looking vision is shared of federated learning as well as digital twin integration along with edge AI effectively creating an emerging landscape for future microgrid intelligence, which lays out a framework and roadmap for researchers and practitioners alike. |
