In this paper, an accurate method is presented for fault locating in transmission lines by using a simple instance-based learning algorithm. Features of patterns are adopted based on harmonic spectra extracted by applying the S-transform on only voltage signals measured at one terminal of line. Due to distinct spectral patterns of transients induced by different types of faults, short circuit faults are divided into the first category including singleline- to-ground and double-line faults, and the second category including double-line-toground, three-line, and three-line-to-ground faults. Then, by relying on flexibility of the utilized learning method, a separate fault-location algorithm is proposed for each fault category. Furthermore, for the first category of faults, a method is provided to identify voltage signals with insufficient harmonic content for distinguishing between reliable and unreliable fault-location results. The accuracy and performance of the proposed method is validated by fault locating in the New Zealand Lower South Island system.

Keywords