Due　to　their　high　efficiency,　photovoltaic　(PV)　cells　can　power　the　Internet　of　Things　(IoT)　devices,　including　sensors,　actuators,　and　communication　devices.　Generally,　PV　cells　are　connected　in　series　to　obtain　a　greater　voltage　without　losing　energy　and　active　area.　However,　a　series　connection　is　unstable,　and　any　fault　in　an　array　inevitably　leads　to　a　breakdown　of　the　branch　or　even　the　system.　Therefore,　fault　detection　is　essential.　This　study　presents　a　systematic　review　of　stat-of-the-art　fault　diagnosis　methods　(FDMs)　of　PV　cells.　We　categorise,　evaluate　and　summarise　the　fault　detection　methods　into　three　broad　areas:　physical,　threshold　and　artificial　intelligence　(Al)　techniques.　Physical　FDMs　detect　the　faults　by　comparing　the　inner　characteristics　ofPhotovoltaic　(PV)　cells　or　their　derived　parameters　with　the　expected　values.　Threshold　FDMs　compare　the　fault　PV　characteristics　with　the　ones　under　normal　conditions.　AI　FDMs　detect　faults　by　employing　a　trained　intelligent　classifier.　Regarding　the　accuracy,　the　AI　FDMs　achieved　an　accuracy　of　no　less　than　90%,　and　some　of　AI　FDMs　even　obtained　100%　accuracy.　Methods　belonging　to　each　category　are　introduced　in　detail.　Finally,　the　summary　is　given,　and　the　developing　tendency　is　recommended　for　future　work.