High Impedance Fault Detection on Power Distribution Feeder Using Subtractive Clustering Fuzzy System

A novel algorithm based on discrete wavelet transform DWT and subtractive clustering fuzzy inference system is presented to detect high impedance fault. As HIFs detection is usually very difficult using the common over current devices, both frequency and time data are needed to get the exact information to classify and detect no fault from HIF. Discrete wavelet transform is utilized for decomposing typical current, voltage and power of high impedance fault current signals. A specific comparison is made among many types of features of the voltage signal, current signal and power signal. The effect of switching of capacitor bank, switching of no-load line, linear and non-linear load current, and harmonics of other normal event on distribution system is presented. Simulation of a 13.8 kV distribution system using PSCAD were done to obtain the HIF signals and other operation event signals. The proposed method shows that it is more convenient for HIF detection in distribution systems with ample varying in operating cases.