Abstract: With low porosity, low permeability and complex structure, the Permian fractured glutenite reservoir in B31 fault block which is located in the western margin of Junggar Basin is mainly controlled by the structure and lithology as well as formation curvature, etc. The static geomodel of fractured reservoirs is built with stochastic modeling approach for simulation. The author integrates a set of rock matrix properties and fracture-related seismic attributes and production data to simulate a 3D fracture distribution model with fracture indicator, fuzzy logic and neural net techniques. It also shows that the use of quantitative fracture distribution (orientation and probability) estimations, together with their error bars (confidence bounds), is a valuable tool for fracture reservoir characterization and the sweet-spots identification with reduced risks.

Key words: fractured reservoir modeling, fracture indicator, fuzzy logic, neural net, fracture distribution probability