Sample Entropy-Based Reconstruction Model for Low-Voltage Arc Monitoring

Arc faults are a significant risk to electrical safety and grid stability because they can cause electrical fires through open arcs and carbonized paths. This study focuses on accurately identifying arc fault characteristics by collecting high-frequency current and voltage data and analyzing them using sample entropy. To improve the analysis, we have enhanced traditional Variational Mode Decomposition by integrating the Grey Wolf Optimization algorithm to automatically optimize parameters and minimize the mutual information of mode pairs. Additionally, an RBF neural network has been used to model arc faults under various conditions, creating a predictive model for short-circuit arcing and carbonized paths. This model, trained and validated with selected datasets, has shown high predictive accuracy and improved training speed, providing effective early warning capabilities.