The invention discloses a 
flutter boundary prediction method in a wind tunnel 
flutter test under a turbulence excitation condition. At first, 
time domain response signals of typical measuring points in the process of the wind tunnel test are collected, trend removal and smooth pretreatment are carried out on the signals, self-correlation 
processing is conducted on processed turbulence response data, the processed turbulence response data serve as impulse responses of a 
system, band-pass filtering is carried out on the processed turbulence response data, and frequency selected by filtering is determined by dangerous 
modal frequency measured by a model 
ground resonance test; secondly, 
spectral analysis and 
modal parameter identification are carried out on the obtained and processed impulse responses with 
noise, and an MPM with high anti-
noise performance is adopted to identify various order 
modal parameters; at last, frequency and damping ratios of various order modals are arranged according to different wind speeds when data are collected, and a mode that straight 
line fitting and polynomial fitting are combined is adopted to extrapolate a 
flutter boundary. According to the method, the flutter boundary under the turbulence excitation condition can be forecasted, a traditional flutter judgment method is improved, and improvements of accuracy and safety of the test are facilitated.