An Identification Model for Grounding Fault in Distribution Networks Based on D-S Evidence Theory

Single-phase-to-ground fault (SPGF) is the most common fault in distribution networks, significantly impacting system reliability and safety. Accurate SPGF identification enhances fault handling precision. This study first extracts multi-domain features from fault waveforms to form a candidate feature set. Using multivariate analysis of variance (MANOVA), we analyze feature-fault relationships and select discriminative characteristics for diagnosis. Next, we develop a hybrid model combining the extreme learning machine’s (ELM) rapid training and support vector machine’s (SVM) high-dimensional separation capabilities. The Dempster-Shafer (D-S) evidence theory fuses these results into a comprehensive identification framework. Field data validation demonstrates the model’s superior performance over single-algorithm approaches across all benchmarks.