Abstract: The reef reservoir in Permian Changxing Formation ofWubaiti gas field has relatively low matrix porosity and permeability, but the fractures in the reservoir lead to high yield and make the permeability higher. Therefore, the key task in the research of the area is to discriminate and predict the degree of fracture development in single well by using conventional logging data. Through breaking core, the conventional log response eigenvalue of the section where fractures developed is shown. By taking the conventional log response eigenvalue as the sample, the BP neural network model based on conventional log response is established to predict the degree of fracture development in the reservoir. The total fracture density predicted by using this model matches core observation to a great extent, which proves that this model is applicable for the study area and could be used to discriminate and predict the degree of fracture development in single well in the area.

Key words: conventional logging, carbonate, fracture, BP neural network, FMI, total fracture density, Wubaiti gas field