Effectiveness of 4D Seismic Data to Monitor CO2 Plume in Cranfield CO2-EOR Project

Abstract

Using carbon dioxide for enhance oil recovery (EOR) has attracted a great deal of attention as the world grapples with the twin challenges of improving oil recovery from mature oil fields and reducing emissions of greenhouse gases. CO2 is considered to be a favorable EOR agent as it can result in a miscible flood process while being less expensive compared to the other similar choices (solvents). Although very fascinating when expressed in terms of numbers and benefits, the life-saver CO2 flood/sequestration projects may turn into a potential risk to the environment when not operated and monitored appropriately. The fact that many oil and gas formations have hold hydrocarbons for millions of years prior to hydrocarbon production provides the fundamental support for possibility of storing CO2 in similar formations however it may induce the risky conception of over-trusting these formations to securely store CO2. Hence, monitoring the CO2 injection process appears to be a necessity in these projects, not only because it helps to improve the efficiency of the process, but also to reduce the risk of CO2 leaking out of the storage zone. In this paper, we investigate the reliability of 4D seismic interpretations as a potential monitoring tool by comparing its results to that of numerical simulation in a CO2 injection project at Cranfield, Mississippi. The numerical simulation model is tuned using historic pressure trends, oil and gas production rates, and current CO2 production data and its validity is confirmed before using it for predictive simulations. Two seismic surveys were performed before and after CO2 injection has started in Cranfield. The first seismic survey was acquired in 2007 (baseline) and the repeat survey was acquired in 2010, when approximately 2.2 million metric tonnes of CO2 has been injected into the reservoir. Seismic surveys were performed in the northeastern part of the reservoir. The overlapping region covers an area of approximately 6 km × 6 km. Using seismic surveys as monitoring methods has the advantage of covering a large area where it can provide information about the CO2 saturation/distribution at locations in which no monitoring well exist. Three different interpretations of the seismic data are discussed in this project. The first interpretation is achieved by generating the time-lapse seismic amplitude difference maps (comparison of the post-injection and pre-injection surveys). The time-lapse amplitude versus angle (AVA) data is then further analyzed by applying a pursuit pre-stack inversion to obtain elastic properties of the reservoir. The last interpretation that is discussed in this papers is achieved by linking the elastic properties of the media to the variations of CO2 saturation, lithology, and cement content using rock-physics models. The distribution maps achieved from each of the above mentioned interpretations are then compared to the numerical simulation results. According to the results, although sensible match between 4D seismic interpretation and simulation results is observed in areas of higher CO2 saturation, areas with lower CO2 saturation show a less clear match. In fact, our compassion of CO2 saturation distribution with interpreted 4D seismic data is not as successful and there are clear discrepancies. We had access to three independent interpretation of the 4D seismic data and discrepancies are observed both in locations with high and low CO2 saturation in all interpretations. It is clear that 4D seismic interpretations have some issues as they are incapable of interpreting CO2 saturation around the injection wells. It seems that collected 4D seismic data in Cranfield suffer from several challenges, impacting the reliability of the interpretations. These challenges include the depth of the formation (higher pressure makes the CO2 density to be higher reducing its contrast with oil and water), thickness of formation (net sands are less than 20 m), presence of methane in the reservoir (this can hinder interpreting the added CO2), and low gas saturations (relative permeability curves also show that gas saturation would barely go over 30-40% in most part of the reservoir).