An intelligent oscillation mode identification method based on Prony analysis comprises the following steps of step S10, selecting a parameter curve whose an oscillation mode needs to be identified and carrying out the Prony analysis on the curve; step S20, according to a Prony fitting error, calculating an oscillation mode identification floating threshold; step S30, searching a mode which satisfies a certain frequency and damping ratio scope and has a maximum amplitude as a first main oscillation mode in a Prony analysis result; step S40, filtering the oscillation mode whose the amplitude is less than K2 times of a first main oscillation mode amplitude in a Prony result, in residual oscillation modes, using a trial and error mode to delete each mode one by one; step S50, extracting a main oscillation mode in the certain frequency and damping ratio scope and identifying a result. In the invention, the main oscillation mode is reserved and intelligent identification of the oscillation mode is realized. The method possesses advantages that thinking is clear; versatility is good and the method is suitable for popularization.