A new method to predict brittleness index for shale gas reservoirs: Insights from well logging data

Abstract

Brittleness is a significant mechanical property which determines the degree of fracturing difficulty in shale gas exploitation and hence, the brittleness index (BI) is a key parameter for the selection of suitable hydraulic fracturing intervals in shale gas reservoirs. However, at present there are no universal and convenient methods for the prediction of BI. Therefore, this study proposes a well logging data-driven BI prediction method based on principal component analysis (PCA) and back-propagation neural networks (BPNN), which can interpret nonlinear and complex relationships between predictors (input variables) and predicted values (target variables). Specifically, 63 samples and their corresponding well logging data were collected from the Wufeng-Longmaxi and Baota formations at the periphery of Sichuan Basin. Experimentally measured brittleness index (BIcore) values (from 0.18 to 0.88) derived from compression test stress-strain curves, were considered as the target parameter. Based on the results of linear regression and sensitivity analysis, five types of well logging, including gamma ray (GR), density (DEN), compressional wave slowness (DT), neutron porosity (CNL) and spontaneous potential (SP), were selected as appropriate well logging techniques. Three principal components were extracted from the five well logging datasets using PCA. Following PCA, the back-propagation neural network (PCA-BPNN) model exhibited excellent prediction accuracy according to statistical and graphical error analysis. The established PCA-BPNN model was then applied to forecast the BI profiles of the intervals of interest in well Y1 in the study area and the predicted BI values were found to be in agreement with the laboratory measured BIcore values. Therefore, the established PCA-BPNN BI prediction method can be applied to accurately locate suitable hydraulic fracturing intervals for economic and technical practicality.