A　method　based　on　wavelet　denoise　and　improved　Prony　algorithm　were　proposed　to　identify　low　frequency　oscillation　modes,　in　view　of　the　fact　the　Prony　algorithm　is　very　sensitive　to　noise　and　high　requirements　for　the　input　signal.On　the　basis　of　wavelet　denoising　analysis,　the　threshold　technique　was　improved,　and　the　wavelet　threshold　was　made　changing　with　wavelet　decomposition.Then　an　improved　Prony　algorithm　were　researched　to　recognize　the　mode　of　low-frequency　oscillation.Through　building　signal,　IEEE　4　machine　2　area　system　simulation　and　experiments,　the　proposed　wavelet　denoising　method　and　Prony　algorithm　were　proved　to　have　a　high　accuracy　and　good　robust　performance.It　is　very　feasibility　to　deal　with　the　noise　of　power　system　and　can　identify　low　frequency　oscillation　modes　rapidly　and　accurately　in　a　certain　noise　environment,　and　provide　an　effective　method　to　power　system　low-frequency　oscillation.　©　2017,　Harbin　University　of　Science　and　Technology　Publication.　All　right　reserved.