Arc　fault　is　an　important　cause　of　electrical　fire　in　distributed　power　generation　and　smart　grid,　which　has　caused　significant　property　losses　and　personal　injuries.　Traditional　protective　devices　such　as　circuit　breakers　and　fuses　cannot　effectively　detect　low-current　arc　faults,　so　it　is　necessary　to　develop　fast　and　accurate　detection　methods.　Therefore,　this　paper　proposes　an　integrated　multi-feature　arc　fault　detection　method,　which　combines　time-domain　and　frequency-domain　features　and　uses　a　multi-level　discrimination　method　to　improve　detection　accuracy.　The　proposed　method　includes　using　numerical　calculations　and　fast　Fourier　transform　algorithms　to　extract　key　features　from　the　current　waveform,　such　as　waveform　factor,　kurtosis,　fundamental　frequency　harmonic　ratio,　and　spectral　centroid,　for　comprehensive　analysis　to　improve　the　accuracy　of　fault　detection.　Its　uniqueness　lies　in　the　use　of　feature　complementarity　and　volatility　to　enhance　fault　recognition　under　complex　load　conditions.　The　integrated　multifeature　detection　device　implemented　using　the　STM32　microcontroller　was　tested　on　an　arc　fault　testing　system.　The　results　show　that　the　proposed　method　has　good　detection　effects　under　various　load　scenarios,　and　its　response　time　in　all　tests　included　in　the　standard　is　far　below　the　specified　threshold.　The　proposed　algorithm　significantly　improves　the　accuracy　and　response　speed　of　arc　fault　detection　and　further　reduces　hardware　complexity　for　practical　application.　©　2025　IEEE.