Data-Driven In-Situ Sonic-Log Synthesis in Shale Reservoirs for Geomechanical Characterization
Conference Paper
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
SummaryCompressionaltraveltime (DTC) and sheartraveltime (DTS) logs acquired using soniclogging tools are crucial for subsurface geomechanical characterization. In this study, 13 easytoacquire conventional logs were processed using six shallowregressiontype supervisedlearning modelsnamely, ordinary least squares (OLS), partial least squares (PLS), ElasticNet (EN), least absolute shrinkage and selection operator (LASSO), multivariate adaptive regression splines (MARS), and artificialneural 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 shallowlearning models are trained and tested with 8,481 data points acquired from a 4,240ftdepth 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 1,460ftdepth interval. After that, five clustering algorithms are applied on subsamples of 13 easytoacquire logs to identify clusters and compare them with the logsynthesis performance of the shallowlearning models. A dimensionalityreduction algorithm, tdistributed stochastic neighbor embedding (tSNE), is used to visualize the petrophysical/statistical characteristics of the clustering algorithm. Hierarchicalclustering, densitybased spatial clustering of application with noise (DBSCAN), and selforganizingmap (SOM) algorithms are sensitive to outliers and did not effectively differentiate the input data into consistent clusters. A Gaussianmixture model can differentiate the various formations, but the clusters do not have a strong correlation with the performance of the logsynthesis models. Clusters identified using the Kmeans method have a strong correlation with the performance of the shallowlearning models. By combining the shallowlearning models for log synthesis with the Kmeans 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 shallowlearning models during deployment in new wells.
