World CCUS Conference 2025

Large-scale injection of supercritical CO2 into saline aquifers is an analogous process to the disposal of large volumes of saltwater co-produced with oil and gas (SWD). Large-scale SWD in saline aquifers over the past 15 years in the mid-continent of North America has caused a remarkable increase of intraplate seismicity. Since 2010, approximately 3.5 billion m3 of produced water has been injected into sedimentary layers that are in hydraulic communication with seismogenic geologic basement in Oklahoma and Texas alone, triggering more than 4,000 M ≥ 3.0 earthquakes. If large-scale CO2 storage projects are associated with induced seismicity, they could be perceived by the public as a hazardous activity which should not be allowed to continue. Earthquake sequences in crystalline basement are frequently triggered by very small pressure increases in overlaying injection zones, frequently about 1 MPa. Injection into sedimentary formations in hydrologic communication with faults in underlying seismogenic basement is responsible for all relatively large (M≥4) induced earthquakes associated with SWD to date and CO2 injection into such formations must be avoided.

This study presents an innovative approach to integrating machine learning (ML) with reservoir simulation to enhance the predictive accuracy and efficiency of CO2-enhanced oil recovery (CO2-EOR) and storage in the SACROC field of the Permian Basin. Traditional reservoir simulation methods are computationally expensive and time-consuming, particularly when dealing with complex geological formations. By leveraging ML algorithms—Random Forest, Gradient Boosting, and XGBoost—this study aims to provide faster and more accurate predictions of oil and CO2 saturation distributions. A dataset of 13,600 data points, including key reservoir parameters such as permeability and porosity, is used to train the models. Performance evaluation through statistical metrics such as RMSE, MAE, and R² identifies Random Forest as the most effective model, achieving the lowest prediction error and highest accuracy. The results demonstrate the capability of ML models to accurately map reservoir dynamics over short- and long-term periods, offering valuable insights for optimizing CO2 injection strategies and improving carbon capture, utilization, and storage (CCUS) efficiency. The findings highlight the potential of ML-driven frameworks to support decision-making processes in subsurface engineering, ultimately contributing to sustainable energy development and climate change mitigation efforts.

This paper utilizes the OLGA code, an enhanced commercial flow assurance tool for CO2 applications, to evaluate the potential backflow of CO2 injection wells under different storage conditions. The OLGA simulation accurately predicts the thermal dynamics of the CO2 stream in the Snøhvit CO2 injection well during its shut-in period. The simulation results show that there is no backflow from the reservoir into the CO2 injection well during normal operational conditions, such as the shut-in of the wellhead choke, if the CO2 entering the well has the same temperature as the ambient temperature, regardless of whether the CO2 is single-phase or two-phase flow in the well prior to or during the shut-in period. However, backflow into the well may occur when the CO2 temperature at the wellhead is higher than the ambient temperature. If the flow assurance tool is coupled with a reservoir simulator, the amount of backflow volume and fluid composition can be accurately predicted. Backflow is not expected to be an issue during continuous CO2 injection into the well, but it may occur if the pipeline and well system experience large-scale unstable flow, such as slugging.

This research investigates how small-scale heterogeneities in rock formations affect CO2 storage in subsurface reservoirs, focusing on core samples from Australia’s Otway CO2 storage site. The study was motivated by observations of unexpected rapid CO2 plume migration at storage projects like Sleipner, which were not predicted by prior modelling efforts.

Using medical CT scanning, we imaged steady-state CO2 injection in cores taken from the Otway site. A key finding was that porosity distribution alone couldn’t predict CO2 distribution patterns, as some influential features were smaller than the medical CT scanner’s 0.6mm resolution.

The study identified various types of heterogeneities within a 5m interval of the Otway basin, including fine layers, thick layers, and more complex patterns. These different heterogeneities resulted in diverse CO2 distribution and trapping patterns. Importantly, these variations occurred at scales smaller than typical reservoir model grid sizes. The findings contribute to Special Core Analysis (SCAL) for modeling a 10,000-ton CO2 injection project in the Otway basin, highlighting the importance of incorporating small-scale heterogeneities in reservoir models for accurate prediction of CO2 behavior.

In this challenging era of deciding Carbon Capture & Storage (CCS) sites, it becomes not only important but also critical to use all available data for drilling better wells for injecting carbon into the underground storage. This abstract would summarize a case study where seismic inversion & machine learning property modeling data was used to effectively delineate the tight vs non-tight sands; this resulted in a better development strategy for the injection of CO2 into the field. The integration of machine learning & seismic inversion results resulted in deciding the best area for injection of CO2 and planning the development phase for CO2 injection across the entire project life for underground storage.

The objective of this study was to effectively delineate the non-tight sands through seismic inversion & machine learning for underground CO2 storage, map these non-tight sand intervals throughout the field and recommend new well locations in the cretaceous deltaic depleted reservoir for CO2 storage.

It is also important to mention here that the machine learning application for property modeling helped in creating 1000’s of scenarios with blind testing validation.

The shallow CO2 controlled release experiment has been conducted at the Otway International Test Centre (OITC) in Victoria, Australia. The project carried out the injection of approximately 16 tonnes of gaseous CO2 at a depth of ∼77–87 m in the vicinity of the near-surface strike-slip sub-vertical Brumby’s Fault. The seismic monitoring technique – 4D reverse VSP (RVSP) – was employed to monitor the rapid development and movement of the carbon dioxide plume and obtain the 3D time-lapse distribution of the CO2 during and after the injection. The RVSP method involves installing a dense array of surface geophones around the injection well and includes using a high-frequency (∼ 1kHz) downhole source deployed in a nearby monitoring well. The RVSP approach allows the daily acquisition of a 3D seismic vintage, providing frequent updates on the plume evolution. We present the observation of the CO2 migration patterns using the results of seismic tomography of the RVSP time-lapse data. The seismic monitoring program highlights the effectiveness of the 4D reverse VSP technique for rapid data acquisition, making it well-suited for capturing fast-evolving subsurface processes.

In this review paper, we show how sequential multi-physics trapping processes operate within the storage complex to generally reduce the risks of re-migration and leakage. Using the Invasion Percolation Markov Chain model we show how the probability of these CO2 migration events within a multi-barrier geological system can be quantified.

Considering the possibility of vertical migration pathways, such as via active fracture zones or well-bore damage zones, we identify several geomechanical and geochemical processes which have a tendency to add other ‘multi-physics’ retention processes, especially geomechanical creep and carbonate precipitation. The role of these coupled geochemical and geomechanical processes is complex and an important topic of ongoing research. However, laboratory studies to date suggest that geomechanical creep and carbonate mineral precipitation tend to act slowly but effectively to reduce the risks of vertical migration of CO2 in the subsurface.

To gain better insights into CO2 fluxes along faults and fractures in the Earth’s crust, we are studying natural tectonic degassing in the North Atlantic region, where many natural CO2 fluxes have been quantified. We plan to use these natural analogues as a reference model for quantifying possible future episodic leakage fluxes from CO2 storage facilities.

In a pilot project in Helguvik, Iceland, in-situ CO2 mineral storage is tested for the first time using saline water instead of fresh water for injection in the field scale. We present three geophysical and geochemical techniques that are novel to in-situ CO2 mineral storage site characterization and monitoring. The baseline characterization, employing single-hole electrical resistivity tomography (ERT) and crosshole seismic measurements, revealed decameter-thick basaltic layers and allowed us to interfer the subsurface porosity and permeability distributions. Rock physics modelling predicts significant seismic velocity increases associated with secondary mineral precipitation, suggesting crosshole seismic time-lapse surveys as a valuable monitoring tool. Results of the ERT timelapse measurements show distinct variations between the baseline and the timelapse measurements, indicating potential resistivity changes due to secondary mineral precipitation and fluid substitution. Geochemical monitoring of a Helium tracer confirms that injected water has reached the monitoring well at 100 m distance, whereas the CO2 concentration in the same well shows no significant increase relative to the pre-injection state. This work demonstrates the potential of combined geophysical and geochemical methods for characterizing and monitoring in-situ CO2 mineral storage, highlighting ERT and crosshole seismics as valuable tools.

CO2 -storage, similar to drilling for petroleum production, represents a risk in terms of accident potential related to “blow-out” during drilling and considerable loss of acceptance in society for the storage of CO2 in the underground.

Gradually, by establishing a multitude of wells for CO2 -storage, in-fill drilling in mature CO2 storage regions will be required. Accordingly, there will be occurrences with drilling of CO2 wells in the vicinity or within existing reservoir volumes containing CO2. Thus, new wells will be established near or penetrating in the CO2 conductive layers.

To address and bridge the knowledge gap between petroleum wells and CO2 wells, Equinor, Shell, Eni, Ebn, CNOOC, Transocean, Baker Hughes, Wild Well Control, Sintef, Gassnova and eDrilling have conducted a project, where drilling fluids mixed with CO2 under relevant pressure/temperature conditions are investigated in laboratory experiments/tests.

The knowledge from the laboratory experiments is synthesized into simulation software from eDrilling for safe and efficient planning of such operations. This software includes solubility of CO2 in oil and water-based drilling fluids, as well as the impact of phase transition from liquid to gas (“boiling”) when the mixture of drilling fluid and CO2 (liquid) is circulated to the surface.

Geological CO2 storage is a key strategy in reducing greenhouse gas emissions, but its economic viability is challenged by risks associated with complex CO2 phase behaviour, especially when mixed with impurities present in capture processes, as well as those present in underground reservoirs. These impurities complicate aspects such as injection risks, storage capacity, and safety. The CO2plus project aims to address these challenges by developing models to analyse the intricate phase behaviour during injection and storage, focusing on issues like hydrate formation, Joule-Thomson cooling, and salt clogging.

In this contribution phase equilibrium data for CO2 + Ar, CO, H2, H2S, N2, O2, and SO2 are collected from literature and curated for further model development. The Soave-Redlich-Kwong, Peng-Robinson, Cubic Plus Association, and PC-SAFT equations of state are fitted to the curated data and the model descriptions are compared to those of the newly updated EOS-CG, a model designed specifically for humid gases and CO2-rich mixtures. Overall, the models do comparatively well, but EOS-CG outperforms all models in the mixture critical regions.