ABSTRACT
This paper presents fuzzy C-means clustering and K-nearest neighbor regression-based protection scheme of transmission line. The current and voltage signal at both ends of the transmission line are sampled, which are synchronized with a GPS clock. Using moving a quarter cycle, the current wavelet-approximate coefficients of the current samples are subtracted from the previous cycle approximate coefficients of current samples to obtain resultant approximate coefficients of current. Fuzzy C-means clustering is applied on the resultant approximate coefficients of the current samples, to compute two centroids. The difference of these centroids is computed, to get a centroid difference. The fault index is computed by adding the centroid difference of both terminals. The fault index is compared with the threshold value to detect the fault in the line. The fault location algorithm is proposed using the K-nearest neighbor regression method.
