USING WAVELET TRANSFORM AND NEURAL NETWORKS DETECTION HIGH-IMPEDANCE FAULT

摘要

This investigation proposes a novel analysis method that can discover the potential effect of high-impedance faults (HIF). The proposed method incorporates a new scheme for protecting the both covered conductors and bare conductors overhead distribution feeder. HIF is generally accompanied by the weak arc phenomena between the fallen conductor and the ground, possibly causing a fire hazard or endangering humans. However, conventional ground fault protection schemes and algorithms have difficulty in recognizing HIF. The properties of scaling and translation of wavelet transform can be used to identify the low-frequency stable and high-frequency transient signals. Discrete wavelet transformations (DWT) are initially applied to extract distinctive features of the voltage and current signals, and are transformed into a series of detail and approximation wavelet components. The coefficients of variation of the wavelet components are then calculated. This information is sent to the training neural networks to identify an HIF from the operations of the switches. The simulated results clearly demonstrate that the proposed technique can accurately identify the HIF in the distribution feeder.