Quantifying the Limits for Acceptable Variances between Time-Lapsed Logging and Reservoir Simulations of CO2 Plume Migration

In topic 24.d Other Topics Related to Advanced Computing, Wellbore Integrity, Tight Reservoir Recovery, or Plume Detection, the Department of Energy seeks to enable more extensive understanding of belowground monitoring and tools for implementing regulations. The belowground reservoir of CO2 is the greatest risk for success of any sequestration project and therefore the greatest risk for the additional benefits of electricity generation and/or energy storage. DOE/NETL- 2017/1847 provides a comprehensive discussion of best practices for monitoring geologic storage projects. Best practices for assuring the CO2 plume remains within the predicted area recommend time- lapsed logging (direct measurement) in the wells and seismic acquisition (indirect measurement) over the plume’s areal extent. These data are intended to 1) confirm the accuracy of a simulation model for the plume; 2) refine the aquifer’s petrophysical model (static model) with far-field data from the injector; and 3) identify the loss of areas of research containment of the plume. The recommendation presumes time-lapsed data will be an accurate representation, i.e., low variance, of the actual plume distribution and saturation profile within the reservoir. If this is true, and the static reservoir model is accurately representative, then the simulation results should agree with the data. Unfortunately, data have error bars and simulation predictions will be sensitive to the number and distribution of grid blocks. Variances from the true values have many sources: logging tool calibration and evolution in tool sensitivities, subsurface complexity, choice of petrophysical correlations, analyst interpretation, evolution in logging tool sensitivity, etc. Simulators’ predictive accuracy is limited by the quality of the static model, number of grid blocks and their distribution within the volume of interest. TERRACOH proposes to develop a statistical model for determining the statistical significance of variances between time-lapsed logging data and simulation predictions in Phase I. Multiple, statistical ally generated, time-lapse synthetic logs will be developed for an injector and an offset well using existing logging suites of the Deadwood Formation in North Dakota. The southeastern Saskatchewan CO2 sequestration project, Aquistore2, uses the Deadwood formation. CO2 plume evolution will be modeled using NITEC’s COZView reservoir simulator (developed with DOE/NETL support in 2013) with a detailed, representative static model. The resulting, statistical model will provide a tool for operators to evaluate measurements versus predictions for plume evolution. TERRACOH expects to extend the model to include indirect measurement uncertainties in Phase II incorporating synthetic seismic. Benefits of this model flow to CO2 sequestration operators and to regulatory agencies. Operators and regulatory bodies should be prepared for variances from prediction. They must determine if the variances are sufficient to warrant action. Actions that address variances may be disruptive to operations and costly. Initiating any corrective action should be scientifically justified. This application will provide a quantitative methodology for evaluating variances in time-lapse logging from simulation results. By reducing uncertainties in monitoring interpretation, the industry will be better able to secure financing for project development. The regulatory bodies responsible for overseeing monitoring, likewise, will have a better tool for evaluating the security of the sequestered CO2.