A comparative study on methods for evaluation of lightning impulse parameters

Abstract

The parameters of lightning impulse (LI) can be determined as per IEC-60/IEEE-4 standard, if the shape of the impulse is smooth. However, difficulties arise if the oscillations or overshoot are superimposed on the waveform. If the frequency of such oscillation is greater than 0.5 MHz or the duration of overshoot is less than 1 /spl mu/sec, it is essential to determine the mean curve for evaluating LI Parameters. In this work, two approaches in evaluating the mean curve are extensively analyzed. The first approach is based on multiresolution signal decomposition. This technique uses the wavelet transform, where a dyadic-orthonormal function decomposes the signal into a smooth and detailed version of the original signal. After removing high frequency components from detailed signal, the mean curve of the impulse waveform can be reconstructed. The second approach is based on mathematical curve fitting where smooth or oscillating impulse is represented by a non-linear impulse waveform equation, representing a double exponential function with sine and cosine terms and Levenberg-Marquardt algorithm is used to minimize least mean square error. The merits and demerits of the above two methods are compared and analyzed. Case studies involving the standard lightning impulse (SLI), lighting impulse with front oscillations (LIFO), with peak oscillations (LIPO), impulse with long duration overshoot (LILDO), impulse with front as well as peak oscillations (LlFPO) and impulse with noise (LIN) have been considered. Based on the above analysis, a conclusion is arrived.