Improving the Performance of Serial Arc Detection using VMD and Deep Neural Network

Abstract

Serial arc is one of factors causing electrical fires. Over the past decades, a variety of researches have been carried out to detect arc signals using frequency characteristics, wavelet analysis and statistical features. However, the usage of those features has shown low arc-detection performance. To solve this, we employ variational mode decomposition (VMD) to generate more time-domain signals, from which statistical features are computed from VMD mode signals, providing more informative features. Using a deep neural network (DNN) as an arc classifier, experiments validate that the VMD could improve the arc-detection performance by 4 percent.