--> --> Secondary Porosity Prediction in Complex Carbonate Reefs using 3D CT Scan Image Analysis and Machine Learning

2019 AAPG Eastern Section Meeting:
Energy from the Heartland

Datapages, Inc.Print this page

Secondary Porosity Prediction in Complex Carbonate Reefs using 3D CT Scan Image Analysis and Machine Learning


The Midwest Regional Carbon Sequestration Partnership (MRCSP), has been investigating various reservoir characterization, modeling, and monitoring technologies related to CCUS in conjunction with CO2-EOR operations in multiple depleted pinnacle reef oil fields in Michigan, USA. The reefs have an asymmetrical geometry, complex internal architecture, and lithologic variations which ultimately have a diagenetic overprint. These features strongly affect the reservoir performance for traditional oil and gas production, CO2-EOR, CO2 storage, and natural gas storage. Numerous studies have been conducted to better understand the variability and reservoir controls of the reefs, however no studies have been conducted to characterize and predict secondary porosity for these systems. A methodology was developed to integrate wireline log measured porosity with CT derived secondary porosity and apply machine learning techniques to predict secondary porosity from readily available wireline logs. To accomplish this, dual energy CT scans were collected from six wells in northern Michigan. 3D image analysis was conducted to isolate features of interest and quantify those features in terms of percent volume of rock, sizes and volumes of each feature, and feature type. Associated wireline log data (gamma ray, bulk density, neutron porosity, and travel time) was digitized and quality checked. Depth correlations were conducted to ensure confidence in depth matches between core and wireline log data. Finally, a series of machine learning models were applied to the wireline log and CT scan secondary porosity data including regression tree, random forest, linear model, and neural network. The results showed that random forest model best predicted secondary porosity with a correlation of .96 using all four wireline logs but could still confidently predict secondary porosity if only three logs were available. Measured secondary porosity ranged from 0-16.6% with an average near 2%. When the predictive model was applied to multiple wells without core data, the average predicted secondary porosity was 1.7% and occurred near the top of the formations. The study is part of the Midwestern Regional Carbon Sequestration Partnership (MRCSP) Michigan Basin Large-Scale Injection Project under DOE/NETL Cooperative Agreement # DE-FC26-0NT42589 with co-funding by Core Energy, LLC, and several other partners.