부산대학교 도서관

Pitfalls associated with application of conventional filtering, fast fourier transform (FFT), and discrete Fourier transform (DFT) on signals with time-varying amplitude are first studied. These pitfalls are demonstrated by two numerical examples. The Kalman-filtering algorithm is then illustrated by examples to be well capable of tracking time-varying amplitude. The proposed high-impedance fault-detection scheme uses the Kalman-filtering algorithm for the optimal estimation of low-order harmonic components. A logic reasoning method is then applied to the estimated harmonic components for determining the possible cause of the disturbance. The performance of the detection scheme is evaluated by several staged arcing faults. A full investigation of these actual recorded arcing-fault data is also performed. Analysis results are used to justify the threshold values used in the proposed detection algorithm and to study the trends of variation of low-order harmonic components in the presence of a high-impedance fault.