Customized wavelets for fault location in power systems

This thesis covers a procedure for generating customized wavelets for detecting the location of a fault in a power transmission line. A typical fault waveform consists of a 60Hz component and multiple reflections of a short transient. Measuring the time-difference-of-arrival (TDOA) between two consecutive transient reflections provides an accurate estimate of the location of the fault. Highly accurate TDOA estimates can be obtained only when the processed fault waveform consists of short/peaked pulses. Our customization procedure enhances the "peakedness" of the transient waveform by optimizing a suitable peakedness metric. This approach yields more accurate estimates than the method used in [1], which used a standard (Daubechies-4) wavelet to process the fault waveform. Our analysis also provides us with some insight as to why a discrete wavelet transform (including the one used in [1]) can significantly enhance the peakedness of a wide variety of power system fault waveforms.