Abstract: Aiming at the problem of complex reservoir structure and difficult prediction in Xiaohaotu-Da16 well area, Bayesian mining algorithm and computer-aided simulation method were used to build the fine hydraulic unit model and correlate the development law of reservoir sand bodies in the target area. The results show that the characterization technology based on Bayesian inference can effectively identify the types of reservoir hydraulic units and provide accurate geological information for gas reservoir description. The exponential function increment method performs better in the process of well logging curve segmentation. The clustering rule of the samples in R_t logging interval is attributed to the high water-cut characteristics of the reservoir. The three types of hydraulic units have large original oil and gas reserves and potential, and they are also urgent to be broken through in EGR process. The research results provide reference for predicting favorable reservoir areas.

Key words: Bayesian inference, tight reservoir, hydraulic unit, flow zone index, well logging intersection, unit forecast, dense gas