Transmission line fault type classification based on novel features and neuro-fuzzy system

This article presents an adaptive neuro-fuzzy inference system and a set of novel features for the classification of transmission line fault types. The ten common types of faults, including line-to-ground faults, line-to-line faults, line-to-line-to-ground faults, and three-phase faults, are considered in this research. The proposed method employs only current waveforms, and the new features include correlation coefficients and inter-quartile ranges of current signals. For the decision-making system based on the neuro-fuzzy technique, two schemes have been investigatedone consisting of 128 rules and the other with 10 rules. Evaluation studies based on both electromagnetic transient program simulated data and field data have demonstrated very promising results for the proposed method.