Design and operation of future low-voltage community microgrids: An AI-based approach with real case study

The utilization of artificial intelligence in the design and operation of a microgrid (MG) can contribute to improve its energy efficiency, resiliency, and cost of energy supply. This research proposes a new approach to conduct a comprehensive analysis for transforming existing low-voltage networks into MGs to achieve the net-zero goal by 2050. A data-driven machine learning-based clustering and profiling approach is designed and implemented to extract the data, constraints, and dependencies from the historical data. Furthermore, the constraints and dependencies are utilized for determining the renewable energy sources’ capacity. A Bi-level optimization technique is developed to ensure appropriate coordination of cost and renewable energy source (RES) capacity. A comprehensive analysis is carried out utilizing real historical demand and generation data of an energy community in Australia. Based on the clustered analysis, the consecutive day's data are considered for the analysis. The findings reveal that the proposed microgrids achieve higher renewable RES utilization and lower electricity costs compared to grid-connected systems, with the potential to reduce carbon emissions by up to 98.23% when transitioning from coal-based grid systems to the proposed microgrid system. Additionally, a transformation from a grid time-of-use tariff-based system to the proposed microgrid setup can lead to a cost reduction of 65.45%. These case studies will also assist the researcher in identifying new, potential ideas and industries to accelerate the implementation of remote community microgrids.