Kurtosis-Skewness Scanning and Machine Learning-based Discrimination of Fault Location in Radial Power Distribution Network

For reliable operation, the detection of fault as well as its location are two important challenges to power engineers. This paper presents an approach to focus on judging the location of buses in a power network where faults occurred. A network having radial busfeeder combinations was considered and data are collected from different buses. Then, wavelet decomposition was done. Different coefficients obtained from the results of signal decomposition were scrutinized by their nature of distribution in terms of Kurtosis and Skewness. Then, different machine learning topologies were applied to network signals for discrimination. They are tested with unknown data sets having a different percentage of randomness and compared. One method is found best that shows a high level of accuracy suitable for judgement of the location where faults occurs.