Geologic Characterization for the U.S. SECARB Anthropogenic Test; Combining Modern and Vintage Well Data to Predict Reservoir Properties
Cyphers, Shawna R.; Koperna, George J.
The U.S. Southeast Regional Carbon Sequestration Partnership (SECARB) Anthropogenic Test is a demonstration of CO2 capture from a coal-fired power plant, transport, geologic storage and monitoring technologies. Starting in August 2012, up to 550 metric tons of CO2 has been captured and injected underground per day. Operations will continue for two years and subsurface monitoring will be deployed through 2017.
The injection site is located within the Citronelle field, a large, producing oilfield in Mobile County, Alabama which was developed over 50 years ago. The CO2 injection zone is the lower Cretaceous Paluxy formation, a saline reservoir located at a depth of about 9,400 feet. The Paluxy formation is a fluvial sandstone and siltstone sequence with sandstone units that vary widely in thickness, lateral extent and reservoir properties. The injection zone occurs above the field's producing oil reservoir, the Donovan Sand. As such, the field's wells fully penetrate the injection zone, providing valuable geologic data. This data is composed predominantly of vintage SP/induction well logs that were used to produce baseline structural and thickness maps of the injection zone and, after calibration with regional core data, to infer general reservoir properties. While this information was essential for a the assessment, the limited vertical resolution of the older logs were insufficient in providing necessary detail of the thin, high permeability and porosity flow pathways present in the Paluxy that can influence plume movement.
A neural network approach provides a method for extrapolating modern, newly acquired to vintage, low-resolution logs. The project's characterization well and two injection wells were drilled in 2011-2012, providing an opportunity to collect a modern well log suite, as well as whole and sidewall cores. The logging suite included MRI, spectroscopy (mineralogy), SP, gamma ray, array resistivity, and neutron, sonic and density porosity. To capture lateral and vertical porosity and permeability variations, neural network techniques were used to generate synthetic porosity logs for ten surrounding wells using only vintage log data. The neural network was trained on the three newly drilled project wells where the modern log suite was acquired. This paper demonstrates how this methodology can be used to generate better geological characterizations of heterogeneity and improve reservoir performance (modeling) predictions.
AAPG Search and Discovery Article #90163©2013AAPG 2013 Annual Convention and Exhibition, Pittsburgh, Pennsylvania, May 19-22, 2013