Abstract
Recently it has become a common practice to construct 3D coreflooding simulation model to interpret water displacement experiments conducted under X-ray CT scanning. The unknown grid block parameters i.e. kro/krw and Pc curves are required to be optimized to get reasonable matching with experimental data such as changes of grid block water saturation. In order to evaluate the matching process efficiently a new automated history-matching program has been developed. This program applies Genetic Algorithm to optimize several coefficients for normalized kro/krw and Pc curves for each litho-facies.
Several blind tests were carried out on hypothetical coreflooding models by changing the conditions of velocity and wettability to investigate the degree of accuracy and limitation of the program. The result of the reproducibility of the relative permeabilities was excellent for both water-wet and oil-wet cases regardless the velocity of coreflooding. On the other hand, the degree of reproducibility was not necessarily satisfactory for capillary pressure curves especially in high velocity case.
Sensitivity of the controlling parameter in Genetic Algorithm such as crossover rate and mutation ratio was also investigated. The suitable values are estimated, though no simple trend was found.
The program was finally applied to the interpretation of actual water displacement tests on oil-wet carbonate cores. The program successfully gave a reasonable set of kro/krw and Pc curves for each litho-facies and demonstrated its capability of grid block parameter optimization.
