This　paper　proposes　a　novel　Prony　method　for　identifying　low　frequency　oscillations　in　power　systems　based　on　improved　smoothness　priors　method　(ISPM)　and　second-derivative　method　(SDM),determining　order　selection　in　low　frequency　oscillations　using　wide　area　measurement　system.　First,　high　frequency　interference　and　trend　components　are　removed　rapidly　and　correctly　with　ISPM.　Next,　SDM-Prony　identification　is　performed　using　denoised　signal　to　obtain　dominant　mode　parameters　of　low　frequency　oscillation.　This　method　can　automatically　and　accurately　determine　order　according　to　specific　conditions　for　singular　value　decomposition　without　artificially　selecting　threshold,making　order　determination　self-adaptive.　The　proposed　methodis　applied　to　simulated　signal　and　actual　oscillation　signal　measurements,　and　obtained　results　were　compared　with　those　produced　with　traditional　Prony.　Using　the　proposed　method,　estimated　order　was　closer　to　actual　order　and　fittingaccuracy　difference　was　not　high.　The　proposed　method　was　characterized　with　simple　calculation　and　excellent　anti-noise　performance,　able　to　identify　dominant　oscillation　modes　rapidly　and　accurately.　Simulation　result　showed　that　this　method　has　good　practicability.　©　2016,　Power　System　Technology　Press.　All　right　reserved.