Sistematical evaluation of digital wavelet filter selection for identification system, case study: DC arcing detection system

The Discrete Wavelet Transform (DWT) is a valuable tool for analyzing waves with nonstationary disturbances, yet selecting the most appropriate wavelet type remains a significant challenge. This research presents a systematic evaluation approach to identify the optimal wavelet type, using Series Arc Faults (SAFs) as a case study. The study evaluates several DWT filter techniques, including Haar Level 4, Daubechies (Db) Level 4, Symlet Level 4, and Coiflet Level 4, establishing uniform SAF threshold values for each wavelet type. The analysis of SAF data was conducted using the center of gravity (CoG) method and feature extraction, with results illustrated in a radar diagram. The findings revealed that the Db 4 method exhibited the highest accuracy in detecting SAFs, surpassing the other wavelet types by 4%. This indicates that the CoG and feature extraction methods are effective for determining the optimal wavelet type in wave analysis.