Applications of tunable-Q factor wavelet transform and AdaBoost classier for identification of high impedance faults: Towards the reliability of electrical distribution systems

Abstract

This study presents a novel approach that employs a mixture of the tunable-Q wavelet transform (TQWT) and enhanced AdaBoost to address the issue of high impedance fault (HIF) recognition in power distribution networks. Traditional overcurrent protection relays frequently have lower fault current levels than normal current, making it exceedingly difficult to detect this HIF problem with the necessity to use a quick and effective approach to find HIF problems. Since the TQWT performs better with signals that exhibit oscillatory behavior, it has been utilized to extract special features for the training of the improved AdaBoost model. The procedure is accelerated by calculating the Kourtosis (K) value for each level and selecting the ideal level of decomposition to minimize computing work. Faulted zones are categorized using an enhanced AdaBoost approach. Under normal, noisy, and unbalanced conditions, the recommended approach is applied to an imbalanced 123-bus test system and an IEEE 33-bus test system. The efficiency of the recommended method is also being assessed for imbalanced distribution networks incorporating dispersed generation into real-time platforms. This procedure is quick compared to previous methods since it uses an upgraded AdaBoost classifier and optimal decomposition level.