Tuesday, December 14, 2017 GSA Luncheon
Topic: Application of Predictive Data Analytics and Geostatistics for Multi-scale Shale Facies
Noon Luncheon 11:30-1:00 pm
Data-driven integrated shale facies modeling is important to visualize distribution pattern of different facies, interpret depositional environments, and understand hydrocarbon potential of mudstone formations at multiple scales- core, well, and regional. The Bakken Formation in the Williston basin of North Dakota is chosen as a casestudy. The main objective of this study is to identify different shale facies in the upper and lower Bakken members to investigate their vertical and lateral heterogeneities at multiple scales for better understanding of depositional controls on mineralogy and Total Organic Carbon (TOC).
Exploratory data mining is utilized to classify shale facies based on quantitative mineralogy and TOC from core data (such as XRD, XRF, and pyrolysis) and advanced pulsed neutron spectroscopy logs. After core-based observation and validation of quantitative shale facies, predictive data analytics techniques, such as Support Vector Machine (SVM) and Bayesian Network (BN) theory, are used to learn the pattern of different facies associated with petrophysical properties from ubiquitous conventional well logs from ~500 wells. A set of ten petrophysical properties are used as input parameters to the SVM algorithm, which can directly output shale facies with high
The results show that the upper and lower Bakken shale members are vertically and laterally heterogeneous at core, well, and regional scales, but can be classified into five different facies. Five shale facies possess different petrophysical and geomechanical properties. Organic-rich shale facies is more dominant than organic-lean shale facies. It appears several factors, such as source of minerals, paleo-redox conditions, organic matter productivity, and preservation etc. controlled the Bakken shale facies distribution pattern. Organic-rich siliceous shale facies shows positive correlation with hydrocarbon production.
Shuvajit Bhattacharya, UAA, Anchorage, AK
Dr. Shuvajit (Jit) Bhattacharya is currently an assistant professor in the Department of Geological Sciences at University of Alaska Anchorage. He teaches courses in integrated subsurface mapping, geophysics, and petrophysics. His broad research themes are energy geosciences, quantitative rock property analysis, and big data analytics. Prior to joining UAA, he worked with EOG Resources, Talisman Energy, and Battelle. He completed multiple projects for unconventional energy exploration, enhanced oil recovery, and carbon storage in North America, Australia, and South Africa. He received awards from AAPG and SPWLA for best presentations.