Classification and localization of transmission line faults using curve fitting technique with Principal component analysis features

Abstract

This paper investigates the application of Principal Component Analysis (PCA) for the development of a simple method of classification and localization of power system faults for a 150 km long transmission line. The proposed work uses only a quarter cycle pre-fault and half cycle post-fault receiving end line current signals for fast identification and isolation of the faulty line. This work analyzes current data of ten different fault classes. The fault signals are recorded at fourteen intermediate geometric locations, out of which, three locations are used for developing the PCA-ratio based classifier and a total of six locations are used for developing the localizer model. PCA is applied here to develop PC score indices, based on which, fault signature curve is developed using best fit analogy. This curve works as the key signature of localizer for each class and phase. The work is made more practically suited by incorporating noise in the signals. Thus an effort has been made in the proposed work for developing a complete practical fault diagnosis algorithm with an aim to achieve high level of accuracy both to classify and localize fault. The proposed classifier is found to produces 100% accuracy, and the localizer is found to achieve an average localization error of only 0.1189% for 40 dB SNR and 0.3965% error at a further higher noise level of 25 dB SNR, with less than 4% of maximum error.