High impedance fault identification using FIR filters, wavelet transforms and computational intelligence

Abstract

This thesis deals with a computer-based simulation study of high impedance fault detection on power distribution systems. High impedance faults (HIFs) are difficult to detect. When a conductor such as a distribution line makes contact with a poor conductive surface or substance the resulting level of fault current is usually lower than the nominal current of the system at the fault location. Therefore, conventional protection relay system will not be able to detect the HIFs and trip the protection relay. The failure of HIF detection leads to potential hazards to human beings and potential fire hazards. HIFs on electrical transmission and distribution networks involve arcing and/or nonlinear characteristics of fault impedance which cause cyclical pattern and distortion. Therefore, the objective of most detection schemes is to evaluate the special features in patterns of the voltages and currents in HIFs. Some researchers proposed various detection schemes based on fractal techniques, digital signal processing, expert systems, neural networks, crest factor, wavelet transform in high frequency noise patterns and dominant harmonic vectors. They offer potential solutions to these problems currently associated with conventional schemes. While direct calculation of fractal dimensions is not effective due to relatively short data sets available for estimation, the use of high frequency harmonics is not feasible in practical measurement because of the filtering by substation current transformers. Simulations using the Electromagnetic Transients Program (EMTP) and Matlab were employed to perform a stochastic study of the nature and waveforms of the fault voltages and currents in the AC supply. This thesis proposes a method which incorporates the statistical nature of the high impedance faults and fault locations in order to recognize the characteristic distortions on voltage and current waveform in the electrical supply network. The immunity of the proposed detection to confounding from contingencies such as load and capacitor switching in electrical networks is evaluated through simulation. After capturing the voltage and current waveforms from the power system simulations, they were analyzed by finite impulse response (FIR) filter bank and discrete wavelet transform (DWT) followed by rms conversion to produce rms values under different frequency ranges. The rms values of these voltage and current waveforms in their various frequency ranges were fed into the pattern classifier such as nearest neighbour rule (NNR) method and artificial neural network (ANN) to determine the fault or non-fault situations. The sensitivity test of above simulations was also performed to investigate the lower current HIFs. Three types of power distribution systems were used for the demonstration of the high impedance fault analyses.