Intelligent energy management of coordinated community microgrid systems using metaheuristic optimization and deep learning

This paper presents an intelligent energy control framework for interconnected community microgrids, integrating metaheuristic optimization with deep learning for optimal dispatch in three configurations: (i) independent grid-connected microgrids, (ii) coordinated grid-connected multi-microgrid systems (MMGs), and (iii) islanded MMGs. The unified modelling approach enables consistent evaluation of operational performance, resource utilization, and cost efficiency under identical assumptions. The proposed strategy is validated through simulations in a Python-based environment using one-day operational data (96 × 15-minute intervals) for three community microgrids. Comparative results demonstrate that coordinated grid-connected operation (Case 2) achieves up to 18.77% reduction in total electricity cost compared to independent operation (Case 1), primarily due to effective inter-microgrid power exchange and optimized resource allocation. The GWO achieved the lowest electricity cost (2260.50) in coordinated grid-connected operation, outperforming ABC by 26.75% and the DL framework by 8.44%, with ANOVA and Friedman tests confirming significant performance differences. Although the islanded MMG configuration (Case 3) incurs slightly higher costs, it eliminates dependency on the main utility grid, enhancing system resilience against market volatility and supply disruptions. Operational analysis shows that coordination improves renewable energy utilization, reduces peak grid transactions, and balances battery charging/discharging patterns. In islanded mode, a combination of battery energy storage and inter-microgrid exchange ensures reliable supply–demand balance without external support, demonstrating the feasibility of self-sufficient, resilient MMG architectures. The proposed methodology offers a scalable and adaptive solution for optimizing distributed energy resources in interconnected microgrid environments, supporting both cost minimization and resilience enhancement. This work can be extended to multi-day horizons, integrate dynamic market pricing, incorporate electric vehicle charging strategies, and assess fault-tolerant performance under adverse operating conditions.