Simulated 330 kV Transmission-Line Fault Dataset and Reproduction Code for Multi-Scale Morphological Fault Classification

This record contains the simulated dataset, the MATLAB/Simulink model, and the Python analysis code supporting the paper "A Lightweight and Interpretable Morphological Framework for Transmission-Line Fault Classification" by C. E. Opata and F. U. Ilo. The dataset (raw_waveforms_20kHz.mat) comprises 250 fault scenarios on a 330 kV transmission line, spanning ten fault types, five fault distances, and five fault resistances. Each record contains the three-phase currents and voltages sampled at 20 kHz over a 200 ms window, with the fault applied at 0.06 s and cleared at 0.1 s. The dataset was generated in MATLAB/Simulink using the model (TransmissionLineModel.slx) and generator script (generate_raw_waveforms.m) included here. The Python pipeline (msmm_pipeline.py) reproduces the clean-signal classification results, the confusion matrix, and the noise-robustness sweep reported in the paper, using multi-scale mathematical morphology on the superimposed current component with a LightGBM classifier, benchmarked against convolutional and recurrent neural-network baselines under five-fold stratified cross-validation. The figure script (make_figures.py) regenerates the manuscript figures. Classification results are deterministic under fixed random seeds; training times are hardware-dependent. See README.md for full instructions and requirements.txt for dependencies.