AC series arc fault detection device based on multiple metrics and Random Forest

AC series arc faults in power distribution systems present significant fire risks, necessitating the development of a device capable of precisely detecting such arcs to ensure equipment reliability. The primary challenge involves differentiating fault currents from normal operational currents due to similar amplitudes and diverse types of loads. This research tackles these challenges by initially developing an apparatus to capture arc characteristic parameters. Employing an ARM microcontroller and FPGA, this device samples arc currents and calculates their characteristics. These parameters are then used to train a Random Forest algorithm, which generates a model capable of accurately classifying arc fault types based on these characteristics. This model is subsequently implemented on an ARM microcontroller for experimental validation of the device's accuracy in detecting arc faults. Experimental results affirm that the device proficiently identifies arc fault currents under various load conditions promptly and provides a dependable fire warning mechanism.