Machine-Learning Assisted Framework for Multi-Microgrid System Identification Over Agricultural Production Areas

The transition toward a modernized power grid founded on smart grid technologies requires the accurate identification of potential subsystems through the analysis of distributed generation, energy storage, and load patterns among clustered prosumers in specific regions. This task becomes particularly complex in large-scale, geographically dispersed distribution systems, especially in agricultural areas characterized by seasonally intensive energy demands. For the bottomup development of smart grid applications in farmlands, it is essential to identify multi-microgrid systems as clusters of interconnected microgrids, considering parameters such as feeder line impedance, seasonal variations in power demand, distributed generation availability, and energy storage capacity. Identifying optimal combinations of prosumers facilitates the formation of multi-microgrid structures, enabling enhanced techno-economic and environmental performance while promoting collaborative solutions to grid-related challenges. This study investigates machine-learning-assisted techniques for system identification to address this problem. The results demonstrate the effectiveness of the proposed approach in identifying multimicrogrid systems using a generic CIGRE benchmark distribution network, thereby contributing to the development of data-driven models for complex energy systems in the agri-food sector.