Stacked Ensemble Learning for accurate Fault Diagnosis in AC Microgrid Systems

Microgrids (MGs) are small-scale power systems that encourage the use of Renewable Energy Sources (RES) and they may or may not be supported by the main grid. MG provides resilience and reliability by decentralizing power generation, offering high efficiency and modularity by strategically managing generation as well as distribution creating a sustainable energy future. Surrounded by the advantages MG holds several challenges of fault detection due to bi-directional power flow, dynamic behavior of different energy sources, and varying magnitudes of fault current in different operating scenarios. To overcome these challenges a sturdy methodology is required which can work effectively and efficiently steer through a dynamic environment where power generation, consumption, and grid topology undergo rapid change. The approached methodology involves a novel ensemble stacking approach that uses several base models that take input data, and the output data of the base models act as input to the final meta-classifier model, which in turn results to give the final output. This paper highlights the potentiality of the ensemble stacking method in MG fault analysis application which is tested and validated in IEEE 7 BUS Real Time System and offers insights for improving the accuracy and efficiency of traditional models.