Unified Threshold MMSD Feature Extraction for DC Microgrid Converter Fault Detection

This paper proposes a unified threshold fault detection method for converters based on multiple-mean square deviation (MMSD) feature extraction to enhance the detection and protection capabilities of DC microgrids during converter faults. The MMSD feature extraction method is designed to amplify and highlight the characteristics of the fault voltage signals, and a simple fixed threshold based on MMSD is presented by capturing the mutation of a probability distribution of voltage fault datasets varies across different stages (stable, fault instant, and post-fault stages). Open-circuit (OC) faults of converters in DC microgrids can be detected accurately and rapidly. Compared to traditional converter fault detection methods, the proposed approach offers improved rapidity, reliability, and cost-effectiveness, and its effectiveness and feasibility are verified by simulation results.