Dynamic Processes of CO2 Storage in the Field Learned from Demonstration Projects at Cranfield, Mississippi and Ordos, China

A number of field-demonstration and industrial-scale projects of geological carbon sequestration (GCS) have been conducted in the world over the last two decades. Hydrological-geophysical-geomechanical monitoring at many of these storage sites provide an opportunity for us to rethink the fundamental processes of CO2 storage in naturally heterogeneous formations. However, complete pictures of field phenomena and processes at most of the sites have not been achieved by integrating field monitoring data with site-characterization data. It often takes years to have a clear picture distilled for a field experiment. Without these pictures of what happened in the field, we could not take advantage of the field testing and monitoring to improve our understanding of the dynamic processes of CO2 storage through site-specific numerical modeling. In this talk, I will present (1) multiscale and multipath channeling of CO2 flow in the hierarchical fluvial reservoir at Cranfield, Mississippi (Zhou et al., 2020a) and (2) thermal fracturing and self-propping induced by liquid CO2 injection into the multilayered reservoirs at Ordos, China (Zhou et al., 2020b).

At Cranfield, Mississippi, CO2 was injected through injection well F1 into the Lower Tuscaloosa Formation at three step rates from 2.92 to 8.27 kg/s from December 1, 2009 to September 7, 2010. The total CO2 injection of 126,246 metric tonnes was followed by a shut-in from September 7 to 28, 2010, when the two monitoring wells (F2 and F3) were killed. I will present a consistent picture of dynamic channeling, invasion, spreading, and breakthrough (CISB) of supercritical CO2 in the hierarchical fluvial reservoir after ten years of integration and analysis of complementary field-monitoring and characterization data. The dynamic CISB with small-scale CO2-flow channels in the F1-F2-F3 cross section was imaged by daily electrical resistance tomography (ERT) and time-lapse crosswell seismic surveys. One, three, and four CO2-flow channels logged at F1, F2, and F3 respectively were dynamically connected with strong temporal variations in CO2 saturation during 221 days of drainage and 81 days of imbibition. Three intermediate-scale CO2-flow channels (with highest CO2 saturation) normal to the cross section were ERT-imaged during late-time drainage. A large-scale, sinuous fluvial CO2-flow channel was imaged by repeat surface seismic survey at the end of the imbibition. The fluvial sandstone channel sinuously bypasses the F1-F2-F3 cross section in a point bar, but the channel is connected to the cross section through an intermediate-scale sandstone channel, forming a complicated flow-channel network. The multiscale flow-channel network (in the fluvial-channel-point-bar system) revealed from the observed CISB enables us to consistently interpret the hydrological monitoring data of three tracer tests, each conducted during an injection-rate step, and pre-injection hydraulic-thermal-tracer tests. This interpretation of the CISB and flow-channel network can guide future modeling and data inversion to best understand the effects of natural heterogeneity on CO2 storage efficiency and residual trapping.

At Ordos, China, liquid CO2 at temperature from -15 to 5 ºC was injected into 21 injection layers of five low-permeability formations stacked over 763 m for nearly four years (from May 2011 to April 2015), leading to a maximum bottomhole-temperature reduction of 30 ºC. A unique step-rate injection test with shut-ins was performed annually, with flowmeter and pressure/temperature logging at the end of each injection-rate step. Field observations showed (1) enhanced injectivity with continuous reduction in wellhead and bottomhole injection pressure, (2) instantaneous formation breakdown signaled in high-frequency pressure and temperature transients, and (3) dynamic changes in the feed zone with the highest fractional CO2 flow from a depth of 1920 m to 1750 m and finally stabilizing at 1690 m. We interpret that the two feed-zone changes coincided with thermal fracturing of the second uppermost injection layer, with the pressure transient typical of formation breakdown and fracture propagation, at 13.75 days and of the uppermost injection layer at 386 days. The thermal fracturing was initiated when the total stress change (i.e., pressure increase plus dominant cooling-induced thermal stress) exceeded 94% of hydrostatic pressure. The initiated fractures propagated slowly during CO2 injection, with their thermal plumes retarded by fracture-matrix heat exchange, while they remained self-propping during shut-ins because of the cumulative cooling and contraction of the rock matrix. We attribute the enhanced injectivity to the dramatic system changes caused by thermal fracturing and the subsequent, slow system evolution caused by retarded fracture propagation. The beneficial impact of thermal fracturing is enhanced CO2 injectivity. The negative impacts include CO2 flow preferentially entering thin injection layers that fractured, and reduced storage efficiency in the planned thick storage system with five storage formations and 21 injection layers.

The CISB revealed at Cranfield shows the complex dynamic processes of CO2 storage in naturally heterogeneous reservoirs. The channelized CO2 flow through high-permeability sandstone channels is coupled with CO2 invasion, spreading, and breakthrough, leading to local CO2 storage and trapping in relatively low-permeability lenses of sandstone. The thermal fracturing imaged at Ordos calls for the attention of significant reservoir cooling and thermal stress near injection wells that are often ignored in the GCS community, in comparison with pressure buildup and caprock fracturing.