To　recognize　partial　discharges(PD),　five　kinds　of　typical　defects　in　oil-paper　insulation　are　constructed　and　measured　with　current　pulse　method,　and　chaos　method　is　used　to　researched　the　time　series　of　PD　signals.　The　results　reveal　that　the　PD　is　of　obviously　chaotic　characteristic,　and　the　PD　process　is　chaotic　one.　The　PD　patterns　can　be　qualitatively　analyzed　and　recognized　by　the　chaotic　time　series　of　PD　and　their　chaotic　attractors.　Phase　space　reconstruction　parameters　and　post-reconstruction　chaotic　characteristic　quantities　can　be　selected　to　quantify　the　PD　chaotic　characteristics.　The　verification　and　comparison　between　pattern　recognition　effects　of　PRPD　and　CAPD　are　performed　respectively　by　adopting　the　neural　network　of　radial　basis　function　(RBF),　it　is　found　that　both　schemes　possess　good　own　advantages.　The　statistical　operators　in　PRPD　mode　and　chaotic　characteristic　quantities　in　CAPD　mode　are　comprehensively　selected　as　the　input　vectors　of　neural　network,　and　the　average　recognition　rate　can　reach　95%　to　greatly　improve　the　recognition　on　PD.