Storage Capacity Estimation of Commercial Scale Injection and Storage of CO2 in the Jacksonburg-Stringtown Oil Field, West Virginia

Abstract

Geological capture, utilization and storage (CCUS) of carbon dioxide (CO2) in depleted oil and gas reservoirs is one method to reduce greenhouse gas emissions with enhanced oil recovery (EOR) and extending the life of the field. Therefore CCUS coupled with EOR is considered to be an economic approach to demonstration of commercial-scale injection and storage of anthropogenic CO2. Several critical issues should be taken into account prior to injecting large volumes of CO2, such as storage capacity, project duration and long-term containment. Reservoir characterization and 3D geological modeling are the best way to estimate the theoretical CO2 storage capacity in mature oil fields. The Jacksonburg-Stringtown field, located in northwestern West Virginia, has produced over 22 million barrels of oil (MMBO) since 1895. The sandstone of the Late Devonian Gordon Stray is the primary reservoir. The Upper Devonian fluvial sandstone reservoirs in Jacksonburg-Stringtown oil field, which has produced over 22 million barrels of oil since 1895, are an ideal candidate for CO2 sequestration coupled with EOR. Supercritical depth (>2500 ft.), minimum miscible pressure (941 psi), favorable API gravity (46.5°) and good water flood response are indicators that facilitate CO2-EOR operations. Moreover, Jacksonburg-Stringtown oil field is adjacent to a large concentration of CO2 sources located along the Ohio River that could potentially supply enough CO2 for sequestration and EOR without constructing new pipeline facilities. Permeability evaluation is a critical parameter to understand the subsurface fluid flow and reservoir management for primary and enhanced hydrocarbon recovery and efficient carbon storage. In this study, a rapid, robust and cost-effective artificial neural network (ANN) model is constructed to predict permeability using the model's strong ability to recognize the possible interrelationships between input and output variables. Two commonly available conventional well logs, gamma ray and bulk density, and three logs derived variables, the slope of GR, the slope of bulk density and Vsh were selected as input parameters and permeability was selected as desired output parameter to train and test an artificial neural network. The results indicate that the ANN model can be applied effectively in permeability prediction. Porosity is another fundamental property that characterizes the storage capability of fluid and gas bearing formations in a reservoir. In this study, a support vector machine (SVM) with mixed kernels function (MKF) is utilized to construct the relationship between limited conventional well log suites and sparse core data. The input parameters for SVM model consist of core porosity values and the same log suite as ANN's input parameters, and porosity is the desired output. Compared with results from the SVM model with a single kernel function, mixed kernel function based SVM model provide more accurate porosity prediction values. Base on the well log analysis, four reservoir subunits within a marine-dominate...