Key points are not available for this paper at this time.
• Conducts an in-depth analysis of state-of-the-art nature-inspired and multi-agent-based centralized and decentralized approaches for optimal scheduling of energy generation sources. • Identifies common challenges like premature convergence, and local optima trap in some algorithms. Suggests the optimized parameter tuning, careful adjustments, and hybrid techniques. • Provides a thorough discussion of different ML and DL techniques, and their roles in load forecasting, renewable energy prediction, and weather forecasting. • Highlights the effectiveness of integrated CNN and LSTM architectures in managing the spatiotemporal characteristics, significantly improving forecasting accuracy. Technological advancements, urbanization, high energy demand, and global requirements to mitigate carbon footprints have led to the adoption of innovative green technologies for energy production. The integration of green technologies with traditional grids offers huge benefits. This amalgamation may bring a power mismatch dilemma due to intermittent renewable energy production and nonlinear energy consumption patterns which can affect the whole system’s reliability and operational efficiency. An efficient Energy Management System (EMS) is essential to deal with uncertainties associated with renewable energy production and load demand while optimizing the operation of distributed energy generation sources. This state-of-the-art review presents artificial intelligence-based solutions to improve EMS, focusing on optimal scheduling of generation sources, forecasting load and renewable energy production, and multi-agent-based decentralized control. The review’s finding suggests that the advanced metaheuristic algorithms can overcome challenges of trapping in local optima and premature convergence and due to this, they are now widely adopted and effectively utilized in scheduling problems. To mitigate uncertainties of renewable energy production and load demand, the long short-term memory and convolutional neural networks can manage spatiotemporal characteristics of renewable and load datasets and forecast highly accurate results. The multi-agent-based system offers a distributed control to complex problems that are computationally less expensive and outperforms centralized approaches. The increased use of advanced metaheuristic optimization techniques and hybrid machine learning and deep learning models is observed for optimization and forecasting applications. The advanced metaheuristic algorithms are a good addition to the literature, they are still in emerging stages and their performance can further be improved. This review also presents the decentralized and centralized EMS-based energy-sharing mechanism between interconnected MGs. The use of advanced forecasting and metaheuristic algorithms can potentially handle the stochastic nature of renewable energy production and load demand.
Judge et al. (Tue,) studied this question.