The Influence of Time-Frequency Transforms on AC and DC Arc Fault Signal Representation

Abstract

Series arc faults and system transitions display different electrical signal profiles in multiple kinds of systems. To guarantee effective arc fault detection, the representation influence of time-frequency transform methods is urgent to be understood for the definition of the most promising features to be used for the arc fault detection in ac and dc systems. At first, an experimental setup of arc faults is designed in ac and dc systems defined in UL1699B and IEC62606 standards in this paper. Then series arc fault electrical signals are obtained at different sampling frequency bands. Arc fault noise would show insignificant variations with external noise from operating loads, resulting in detection difficulties for wavelet based methods. Next, time-frequency transform methods including singular value decomposition (SVD) and variation mode decomposition (VMD) are applied to analyze the time-frequency representation of arc fault signals, which would improve the wavelet feature by 8.6% and 30.43% respectively. Finally, the calculation order and Teager energy operator (TEO) are further applied to improve ac and dc arc fault features. This research is aimed to provide portable arc fault features for wider arc fault conditions based on the common arc fault characteristics.