The invention provides an angular part superposition seismic data-based fluid identification method, which comprises the following three steps of: 1) rapidly estimating an impedance reflection coefficient, namely starting from Shuey approximation (Shuey, 1985) in a Zoeppritz equation, deducing a formula for accurately and rapidly estimating the impedance reflection coefficient of longitudinal and shear waves, and calculating a seismic attribute used for fluid identification; 2) extracting a new fluid identification attribute based on the angle gather part superposition seismic data, namely starting from a fluid factor formula provided in Smith and Gidlow (Smith, 2003), and obtaining the combination of the angle gather part superposition seismic data for fluid identification and reservoir prediction; and 3) performing a curvelet transform and Bayesian theory-based fluid identification, namely equalizing each angle superposition gather data by using a wavelet equalization method on the basis of the steps 1) and 2) so as to remove difference of different angles and keep the difference of different angle gather data caused by the fluid, obtaining abnormality at a target layer by using Bayesian wave field separation technology in a curvelet domain, naming the abnormality as angle fluid gather, and solving various attributes to obtain the abnormality at the target layer, wherein the attribute obtained by employing various combinations of the angle gather part superposition seismic data can better distinguish gas/oil-containing and water-containing reservoirs, particularly the attribute section obtained from the third power of the angle gather can better distinguish the gas/oil-containing and water-containing reservoirs.