Empowering DC Microgrids: Detection of Faults With Lanczos Filter and Skewness Analysis

This paper presents a novel communication-free method for fault detection and classification in DC microgrids (DCMGs) using the Lanczos low-noise filter and Bowley's skewness coefficient. In the proposed method, the current signals from each pole are initially processed through a unique feature of the Lanczos filter, a mathematical filter used in signal processing, particularly for reducing noise and preserving important signal characteristics. In normal conditions, the output feature of the first step remains close to zero; however, any variation in the current signals makes it change rapidly. The first difference function is established to enhance the analysis of DCMG fluctuations, which compares each sample with its preceding value. This approach simplifies the diagnostic process for subsequent stages and incorporates Bowley's coefficient of skewness to improve the accuracy of the protective measures employed by the system. This distinctive feature acts as an index for detecting faults. The use of this index enables rapid classification of pole-to-ground (PG/NG) and pole-to-pole (PN) faults under high-resistance (up to 30 Ω) and noisy conditions. The topology-independent method functions seamlessly in grid-connected/islanded modes and radial/mesh configurations, while optimised thresholds ensure immunity to load variations, DG outages, and islanding transients. The proposed technique, validated in more than 2600 simulated scenarios and a 12-V small-scale laboratory prototype, demonstrates superior speed, accuracy, and robustness performance compared to existing methods. This study enhances DCMG resilience and safety with a rapid, reliable, and universal protection framework, detecting faults in 0.9 ms (simulation) and 1.5 ms (experiment) with 97.6% accuracy.