Topology-aware fault diagnosis for microgrid clusters with diverse scenarios generated by digital twins

Abstract

Microgrid clusters, characterized by their dynamically changing topologies and flexible operational modes ranging from grid-tied to autonomous functioning, present significant challenges for conventional protective measures. To address these complexities, this paper proposes a novel topology-aware fault diagnosis approach that integrates Message Passing Neural Networks (MPNNs) with a Graph-Lasso-based topology identification method. Our framework is designed to enhance the precision and adaptability of fault diagnosis in microgrid clusters, which are crucial for maintaining system stability. The proposed model demonstrates over 99% accuracy in fault localization and over 97% in fault type recognition, even under varying topological conditions and data loss scenarios. Through the application of digital twin technology, the framework not only improves the diagnostic capabilities but also supports the intelligent operation and maintenance of microgrid clusters, thereby contributing to the overall reliability and quality of their electrical systems.