A New Fracture Model Integrating Geological Anatomy and Geophysical Detection: A Case Study of Tight Gas Reservoir in West Sichuan Basin, China

Exploration and development of unconventional or complex reservoirs often face the challenge related to fractures. For example, the fracture system effects the hydrocarbon accumulation and reservoir production greatly. Due to the complexity and high spatial heterogeneity of fractures, it is difficult to get a comprehensive and accurate understanding of fractures using a single method such as factor analysis. Taking the tight sandstone gas reservoir of Xujiahe Formation in Xinchang, Sichuan Basin as an example, this paper introduces a hierarchical modeling method for constructing the fracture probability model which combines the geological understanding of fracture patterns and the predictions of geophysical fracture distribution. According to the statistical analysis, the fracture occurrence frequency is mainly controlled by the distance to large-scale faults and the content of brittle minerals. The fault distance coefficient model and the quartz content model which represent the rock brittleness are constructed with the constraint of the fault model and lithofacies model. A seismic attribute named Fault Likelihood is selected due to the capability of fracture system detection and the higher correlation with effective fractures (the ones whose dip angles greater than 30°). It provides us the information of fracture intensity and can be taken as a kind of constrained data for fracture probability modelling. Then we establish the final fracture probability model by weighting it according to the drilling correlation. The Bayesian PR multiple probability fusion is used to form a quantitative fracture probability model combining geological priori and geophysical predictions. Furthermore, a discrete fracture network (DFN) and the corresponding fracture properties model are established with the constraint of the quantitative probability model. The results show that the fracture probability model established by our method not only reflects the probability of rock discontinuity but also denotes the effects of the geological factors affecting fractures. The accuracy of fracture probability modeling result at the locations of boreholes is improved to over 85%. Moreover, the DFN model is verified by the encounter with fractures of newly drilled wells. Hence the result models are the better ones to describe the spatial distribution of fractures. It proves that the modeling method provided in this paper is more suitable for the reservoir characterization and nature fracture prediction especially in tight sandstone gas reservoirs.