This　paper　proposes　a　method　to　localize　multiple　partial　discharge　(PD)　sources　in　air-insulated　substation　by　using　radio　frequency　(RF)　antenna　sensors　combining　with　clustering　algorithm,　which　solves　the　problem　that　multiple　PD　or　interference　sources　are　difficult　to　localize　in　air-insulated　substation　(AIS).　Our　research　group　made　four　omnidirectional　RF　antenna　sensors,　the　detection　frequency　band　was　250　MHz-3　GHz,　the　voltage　standing　wave　ratio　(VSWR)　of　the　RF　antenna　in　the　detection　frequency　band　was　less　than　2,　which　meets　the　detection　requirements.　The　scenes　of　multiple　PD　sources　in　AIS　were　simulated　in　the　laboratory,　the　time　difference　of　arrival　(TDOA)　was　collected　through　four　RF　antennas,　and　the　clustering　effects　of　K-means　and　Two　Step　algorithms　were　compared.　Based　on　the　normal　distribution　characteristics　of　TDOAs　and　the　better　Two　Step　algorithm,　a　3σ　-Two　Step　algorithm　was　proposed　to　process　TDOAs　and　output　the　centers　of　TDOAs.　Newton　iterative　algorithm　was　used　to　solve　the　localization　equations.　The　results　show　that　the　average　localization　error　of　two　PD　sources　was　0.21　m　and　that　of　three　PD　sources　was　0.45　m.　The　PD　localization　tests　were　carried　out　in　an　actual　AIS　in　China,　the　proposed　method　localized　a　corona　discharge　on　a　post　insulator　and　a　corona　discharge　on　a　disconnector.　Through　the　verifications　with　an　ultraviolet　camera,　it　was　obtained　that　the　localization　errors　at　each　position　were　within　the　range　of　0.6-2　m.　©　2001-2012　IEEE.