Texaco has been operating a CO2 flood in the Sundown Slaughter Unit in Hockley County, Texas since January of 1994. The CO2 flood was originally justified by analogy with an adjacent CO2 flood. A CO2 flood simulation was later done to predict and optimize the performance of the flood. With actual production data from the CO2 flood available, the simulation forecast was redone and updated. Specific objectives of this new, revised simulation study were to use geostatistical reservoir characterization to improve the representation of reservoir heterogeneity and to use more representative relative permeability curves and residual saturation values.
A team was formed for the new simulation study which included both geologists and engineers with members both from the operating division and Texaco's research organization. Four new geostatistical reservoir models were developed, each with a different level of effective heterogeneity. The basic idea was to adjust the reservoir characterization to improve the CO2 flood match and forecast. All the new models as well as the old model could be used to match the waterflood history equally well with moderate adjustments in the water-oil relative permeability curves. The correct level of reservoir heterogeneity was not needed to do a waterflood match. However, all the models were not equally valid in matching and predicting CO2 flood performance. The predicted CO2 flood performance was substantially different for these models and indicates that a good waterflood history match is not sufficient for a good CO2 flood prediction.
Having actual CO2 flood production response data proved to be the key factor in choosing a model with the correct level of heteogeneity and generating an improved CO2 flood forecast. A successful match of the actual CO2 flood response could not be obtained with the original model but could be with two of the new geostatistical models. The reservoir heterogeneity in a model must be substantially correct for a successful CO2 flood match. The predicted CO2 flood production response was initially fairly close to the actual for these two models, but the match was improved with adjustments of the gas relative permeability curve. In addition, adjusting the gas relative permeability curves to match the initial CO2 flood response brought the ultimate tertiary oil and CO2 production forecasts much closer than they had been initially. Adjusting the gas relative permeability curve for a CO2 flood history match can compensate for moderate, but not large, errors in the reservoir heterogeneity.
This paper describes the methods used for conducting the waterflood and CO2 flood history matches, for making the CO2 flood forecast, and for evaluating the different geostatistic realizations. In addition, the important sensitivities of a tertiary CO2 flood forecast to the reservoir description and the gas relative permeability are discussed and quantified.