A Low Frequency Source of Transformer Localization Method Using the Combination of CLEAN algorithm based on Spatial Source Coherence and Multi-signal Classification Algorithm

Voiceprint recognition technology, as a contactless monitoring method, can be applied to diagnose the abnormal operating status of a power transformer. Effectively achieving power transformer sound source information is extremely important to realize transformer fault recognition. For voiceprint recognition systems, using CLEAN based on spatial source coherence (CLEAN-SC) algorithm makes locating low-frequency sound source signals from power transformers difficult. In this paper, first, low-frequency signals below 1000 Hz and multisource are distinguished using a multiple-signal classification (MUSIC) algorithm. Then, the CLEAN-SC algorithm is further applied to localize sound sources precisely. Simulation and experiment results show that compared to only using the CLEAN-SC algorithm, the proposed algorithm by combining CLEAN-SC and MUSIC algorithms, i.e., CLEAN-SC-MUSIC algorithm, can effectively reduce the positioning error of the voiceprint recognition system to 0.03 m.