Data-Driven In-Situ Sonic-Log Synthesis in Shale Reservoirs for Geomechanical Characterization

Abstract

SummaryCompressional–travel–time (DTC) and shear–travel–time (DTS) logs acquired using sonic–logging tools are crucial for subsurface geomechanical characterization. In this study, 13 “easy–to–acquire” conventional logs were processed using six shallow–regression–type supervised–learning models—namely, ordinary least squares (OLS), partial least squares (PLS), ElasticNet (EN), least absolute shrinkage and selection operator (LASSO), multivariate adaptive regression splines (MARS), and artificial–neural network (ANN)—to successfully synthesize DTC and DTS logs. Among the six models, ANN outperforms other models with R2 of 0.87 and 0.85 for the syntheses of DTC and DTS logs, respectively. The six shallow–learning models are trained and tested with 8,481 data points acquired from a 4,240–ft–depth interval of a shale reservoir in Well 1, and the trained models are deployed in Well 2 for purposes of blind testing against 2,920 data points from 1,460–ft–depth interval. After that, five clustering algorithms are applied on subsamples of 13 “easy–to–acquire” logs to identify clusters and compare them with the log–synthesis performance of the shallow–learning models. A dimensionality–reduction algorithm, t–distributed stochastic neighbor embedding (t–SNE), is used to visualize the petrophysical/statistical characteristics of the clustering algorithm. Hierarchical–clustering, density–based spatial clustering of application with noise (DBSCAN), and self–organizing–map (SOM) algorithms are sensitive to outliers and did not effectively differentiate the input data into consistent clusters. A Gaussian–mixture model can differentiate the various formations, but the clusters do not have a strong correlation with the performance of the log–synthesis models. Clusters identified using the K–means method have a strong correlation with the performance of the shallow–learning models. By combining the shallow–learning models for log synthesis with the K–means clustering, we propose a reliable workflow that can synthesize the DTC and DTS logs, as well as generate a reliability indicator for the synthesis logs to help the user better understand the performance of the shallow–learning models during deployment in new wells.