Lightning Fault Classification on Transmission Lines using Discrete Wavelet Transform and Artificial Neural Network

Abstract

Accurate classification of lightning faults on transmission lines is crucial in identifying the type of fault, whether it is due to shielding failure or back-flashover. This knowledge is essential in implementing a cost-effective and optimized mitigation method to improve transmission line performance. Previous mitigation efforts focused on improving tower footing resistance (TFR), which does not mitigate shielding failure. This study proposes an artificial intelligence approach to recognize, classify, and distinguish between back-flashover and shielding failure based on waveform signatures of disturbance fault recorders (DFR) with a sampling rate of 5kHz. The methods used in this study are discrete wavelet transform (DWT) utilizing wavelet similarity index, and artificial neural network (ANN). The simulation data using these methods demonstrate an 88.9% accuracy rate for the DWT method, while the ANN method achieves an accuracy rate of 97% for back flashover and 100% for shielding failure using signals at 16.7MHz, while for down-sampled data at 5KHz, the accuracy are 93% and 97% respectively.