Common　failures　of　photovoltaic　array　include　aging　of　the　photovoltaic　circuit,　short　circuit,　open　circuit,　and　shadow　blocking,　which　will　affect　the　generation　efficiency　and　component　safety　of　photovoltaic　array.　In　this　paper,　an　on-line　fault　diagnosis　system　of　the　photovoltaic　array　is　proposed.　Hall　sensor　is　used　to　collect　the　output　voltage　and　current　data　of　the　photovoltaic　array.　According　to　the　difference　of　I-V　characteristics　under　different　fault　types　of　photovoltaic　array,　the　Extreme　learning　machine　model　is　used　to　carry　out　on-line　classification　and　evaluation　of　data,　and　the　current　status　of　the　photovoltaic　array　is　diagnosed　and　displayed.　The　system　has　the　advantages　of　fast　data　analysis　speed　and　high　accuracy　of　working　statejudgment.　©　2020　IEEE.