Research on Multi-Load Arc Fault Detection Based on Dual-Tree Complex Wavelet Transform

In residential electricity scenarios, a wide variety of household appliances are present, and multiple loads are frequently operated simultaneously, resulting in a significant increase in current levels. When there is damaged insulation or poor contact in the electrical lines, it can potentially cause arc faults, which generate high temperatures and pose a serious risk of electrical fires, thereby endangering the safety of residents' lives and properties. Firstly, a multi-load arc fault experimental platform was constructed to collect arc fault waveforms under conditions where typical loads operate simultaneously. A feature quantity evaluation index was introduced to compare the detection characteristics of traditional arc faults under single-load and multi-load conditions. The coupling interference between multiple loads weakens the detection features of arc faults, making it challenging to effectively detect them. Subsequently, an arc fault feature extraction method based on the dual-tree complex wavelet transform was proposed to efficiently extract the time-frequency characteristics of arc faults during multi-load operation. Finally, an arc fault detection algorithm based on the bidirectional gated recurrent unit (BiGRU) was developed. The results demonstrated a 49.52% improvement in the feature separation ratio under multi-load operating conditions, with a detection accuracy of 98.15% achieved on the offline dataset, significantly enhancing the detection effectiveness of multi-load arc faults.