Doctor of Philosophy (PhD)
The continuous researvoir model updating is widely used to calibrate reservoir simulation models to production data, but many challenges remain. First, few real field data are available to test the new history matching method, and most of the data sets are synthetic cases. Second, computational cost may be high when using non-Gaussian priors or nonlinear models. Third, with large complex models, the simulation runs and history matching method require huge memory allocations. This dissertation achieves a continuous reservoir model updating workflow with a meter-scale , two-phase flow experiment. Both production and seismic data are collected in the experiment. Because the data are high-frequency sequential data with noise, the EnKF method is used to efficiently integrate them. To better understand the problem, scaling analysis is done on the capillary transition zone. Two new dimensionless numbers are introduced-capillary time and capillary length. We found that for different models, if their capillary time and gravity number are equal, the capillary length would be the same. The scaling analysis results help us find a proper flow rate for the sand tank experiment. Two experiments are conducted to test the workflow and the EnKF method. In the first one, both the production and seismic data are collected and analyzed. The production data have large errors in the flow rate and they are integrated to improve reservoir models using EnKF method. The history matching results are in an acceptable range which demonstrate that even if the observation data has large error, the EnKF method still works. In the second experiment, the errors of flow rate are reduced by measuring manually with a graduated cylinder. Because the data quality are much better in the second experiment, the observations can be matched easily.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Sun, Ting, "Continuous reservoir modeling updating by integrating experimental data using an ensemble Kalman Filter" (2014). LSU Doctoral Dissertations. 285.