AVO fluid inversion technology is used to calculate hydrocarbon probability quantitatively to reduce the uncertainty of AVO analysis. By means of Monte-carlo stochastic simulation algorithm, well logging data can be used to synthesize forward models with various possible types of fluid combination to get distribution template of oil, gas and water. According to current well production data, this distribution template is corrected and tested. Then it is calibrated with the gradient and intercept of the whole 3D seismic data calculated by Shuey's function. After the calibration, the probability of the hydrocarbon fluid is estimated by means of Bayesian theory. Types of fluid and buried depth of the reservoir are the main factors affecting the result of AVO fluid inversion. This paper focuses on AVO fluid inversion to estimate hydrocarbon probability in the target layer in Yong'an area of Subei basin. It shows that this technology can reduce the uncertainty of fluid prediction and increase the drilling success rate.