Energy management of microgrid coalitions considering renewable energy prediction based on machine learning algorithms

This paper proposes a new energy management framework for the optimum scheduling of distributed generation resources within a smart distribution system that contains multiple interconnected microgrids under normal and abnormal operating conditions. First, a systematic management framework is established by defining the functions of various management units throughout the multi-microgrid system. In the suggested two-step approach, each microgrid schedules its own resources based on the developed model, while in the second step, the distribution system operator makes a decision on power transfers among microgrids and allocates the remaining unserved loads from the first step. Due to the inherent variability of renewable resources, an extreme learning machine (ELM) model is applied in order to forecast solar irradiation and wind power. The significant contributions of the work include (1) a new approach to uncertainty modeling of renewable energy sources and consumer loads, using ELM for probabilistic modeling of the fluctuations of renewables and demand; (2) coalition strategy formation in local and global energy trading for microgrids with an objective of cost minimization and profitability improvement; and (3) incorporation of vehicle-to-grid (V2G) systems for better electric vehicle charging management and contribution to system stability due to renewable energy uncertainties. Besides, the proposed framework has been applied and validated on a test system for practical scenarios. The approach demonstrates considerable enhancements in efficiency, resilience, and cost-effectiveness upon multi-microgrid networks under various operating conditions.