Intrinsic versus Learned Robustness in Transmission-Line Fault Classification under Impulsive Noise: Dataset and Code

Simulated 330 kV transmission-line fault-waveform dataset (250 scenarios: 10 fault types × 5 distances × 5 fault resistances; three-phase currents and voltages at 20 kHz, 4001 samples) and the Python analysis code for the companion paper “Intrinsic versus Learned Robustness in Transmission-Line Fault Classification under Impulsive Noise: A Comparative Evaluation” (Opata & Ilo, ESUT). The code reproduces the paper’s train-clean/test-impaired and noise-aware fairness tables under Middleton Class-A impulsive noise, the operating-envelope sweep, and the front-stage ablation. Data under CC BY 4.0; code under MIT. See README.md.