The　problems　of　strict　sampling　conditions　and　low　positioning　accuracy　exist　in　the　traditional　photovoltaic　array　arc　fault　location　method　based　on　electromagnetic　radiation　(EMR)　signal,　accordingly,　we　propose　a　new　arc　fault　location　method　based　on　grid　fingerprint　matching.　Firstly,　the　EMR　signal　of　the　arc　is　acquired　with　a　low　sampling　rate,　and　its　root　mean　square　value　is　extracted　as　the　characteristic　index　representing　the　EMR　intensity.　Then,　BP　neural　network　(BPNN)　is　adopted　to　mine　the　internal　relationship　among　irradiance,　signal　receiving　distance　and　arc　EMR　signal　intensity,　and　a　prediction　model　is　established.　Subsequently,　according　to　the　predicted　distance　between　the　dual-antenna　array　output　by　BPNN　and　the　arc,　the　area　where　the　arc　is　located　is　preliminarily　acquired　by　using　the　triangulation　method.　Finally,　the　photovoltaic　module　in　the　located　area　is　divided　into　grids　to　generate　grid　fingerprint　information,　and　the　center　coordinate　of　the　grid　that　most　matches　the　predicted　distance　and　fingerprint　information　is　taken　as　the　final　predicted　coordinate　of　the　arc　occurrence　position.　The　experiment　results　show　that　the　proposed　algorithm　has　good　positioning　ability　and　adaptability,　and　the　average　absolute　error　of　arc　fault　location　is　0.306　m,　which　is　superior　to　the　EMR　attenuation　model　positioning　method　in　positioning　accuracy　and　economy.　©　2024　Science　Press.　All　rights　reserved.