Evaluation of fault relative location algorithms using voltage sag data collected at 25-kV substations

Source location algorithms, recently proposed in the literature as a means to determine the origin of sacs upstream or downstream from the measuring point, have been compared. Testing has been carried out using 471 records of asymmetrical voltage sags gathered from HV/MV substations of the Catalan power network, in the northeast of Spain. Six different algorithms have been included in the test, and the significance of their attributes has been analysed using a data mining approach. Multivariate statistical theory (MANOVA) has been used in this analysis. The work has been completed by improving the performance of the algorithms applying a machine learning induction algorithm (CN2), designed to extract new classification rules dealing with the combination of the attributes defined by the algorithms analysed. The Study considers phase-to-ground and phase-to-phase faults separately, Copyright (C) 2009 John Wiley & Sons, Ltd.