Optimal model construction for identifying tectonic fractures in tight sandstone reservoirs via conventional logging

ABSTRACT Fractures serve as effective storage spaces and important seepage channels in tight reservoirs and are thus critical to the exploration and development of such reservoirs. At present, single-well fracture identification is realized mainly via imaging logging, array acoustic logging, and conventional logging. In continental reservoirs, the diversity of the main factors controlling fracture development leads to high randomness and heterogeneity. As a result, fracture identification must overcome problems such as weak effective and highly ambiguous logging response signals. Accurately identifying fractures is a critical issue in the study of tight sandstone reservoirs. In this work, data from 13 cores and statistical data from fracture observations obtained during imaging logging of four noncored wells were collected. Accordingly, a fracture identification database containing different types of conventional logging curves and information on layers, fracture development, and well locations was established. Under the constraints of the stratigraphy and locations of fracture development and the properties of fractures determined via core or imaging log observations, a mathematical model with the highest proportion of identified fractures was optimized to identify tectonic fractures through conventional logging. Compared with imaging logging, the rate of fracture recognition of the proposed method was 77.29%, which is significantly higher than those of the range/standard deviation (R/S) method (66.7%) and the trigonometric projection method (64.6%). Therefore, the method provided a very favorable effect in increasing the fracture recognition rate, analyzing tectonic mechanisms, and performing quantitative predictions.