Time-Frequency Methods Of High Impedance Fault Detection In Distribution Network

With low-resistance grounding method of distribution network being popular in China, high impedance fault detection has attracted people's attention. Because high impedance fault doesn't draw enough fault current to operate overcurrent protective devices, it is hard to detect it. It will threaten people's life or even cause a fire hazard if remains undetected.This paper is financially supported by the Visiting Scholar Program of Chongqing University, aiming to do research on the characteristics and detecting method of high impedance faults. First, the expressions of three phase voltage, fault current and neutral point voltage is derived using symmetrical components method, and the characteristics of single-phase-to-ground fault in low-resistance grounding network and compensated network are compared. Second, for detection method research, line current samples during a kind of high impedance fault accompanied with arcing and normal operations are got through simulation. Then,in frequency domain, FFT is used to get magnitude and phase information of fundamental and harmonic current, and a BP neural network is designed to classify the line current samples. Finally, wavelet transform, the new time-frequency domain analyzing tool, is used to help reduce false classification rate of the FFT and BP neural network based classifier.Simulation results show that the FFT and BP neural network based classifier can distinguish high impedance faults from normal operations. Through combination of wavelet transform and the FFT and BP neural network based classifier, detection of normal operations such as capacitor switching, load switching and inrush current can be more accurate.