A Convolutional Neural Network and Feature Selection-Based Method for Arc Fault Location in Household Appliances

To address the challenge of locating series arc faults in household appliances, this paper proposes an arc fault location method based on feature selection and Convolutional Neural Networks (CNNs). A multi-branch, multi-load-type arc test platform is constructed to analyze the characteristics of various household appliances under series arc faults, and an arc fault dataset containing different load types is created. The proposed method extracts fault features using Fast Fourier Transform (FFT), prioritizes key features with the RELIEF algorithm, and then maps the fault features to fault labels using a CNN, enabling precise arc fault location. Experimental results show that this method significantly improves the accuracy of fault location and addresses issues in traditional methods, such as inaccuracies due to noise interference or similarities in load characteristics.