US 10,890,688 B2
Method for generating secondary data in geostatistics using observed data
Kyungbook Lee, Daejeon (KR); Hyun Suk Lee, Daejeon (KR); Won Suk Lee, Daejeon (KR); Taehun Lee, Sejong (KR); Jungtek Lim, Gyeonggi-do (KR); and Jonggeun Choe, Seoul (KR)
Assigned to KOREA INSTITUTE OF GEOSCIENCE AND MINERAL RESOURCES, Daejeon (KR)
Filed by Korea Institute of Geoscience and Mineral Resources, Daejeon (KR)
Filed on Aug. 30, 2016, as Appl. No. 15/251,253.
Claims priority of application No. 10-2015-0163150 (KR), filed on Nov. 20, 2015.
Prior Publication US 2017/0146690 A1, May 25, 2017
Int. Cl. G01V 99/00 (2009.01)
CPC G01V 99/005 (2013.01) [G01V 2210/665 (2013.01)] 3 Claims
OG exemplary drawing
 
1. A computer-implemented method for generating and using soft data in geostatistics using observed data, the method comprising:
receiving, via a computer, prepared spatial correlation data, hard data, and observed data;
generating, via the computer, initial models by performing a geostatistical technique using the spatial correlation data and the hard data;
extracting, via the computer, a best representative model from the initial models using the observed data and a distance-based clustering method and then, selecting, via the computer, candidate models surrounding the best representative model using the distance-based clustering technique;
determining via the computer, final models after determining that a mean model of the candidate models has converged, but otherwise setting the mean model as the soft data and then repeating the foregoing sequence of steps using the soft data in addition to the prepared spatial correlation data, the hard data, and the observed data; and
after determining a final model, performing, via the computer, dynamic simulation using the final models to estimate uncertainty quantification of future reservoir performance, wherein,
(a) extracting the best representative model comprises (1) forming, via the computer, a plurality of clusters by grouping similar models among the created initial models using the distance-based clustering technique, (2) selecting, via the computer, respective representative models for the plurality of clusters, (3) performing, via the computer, dynamic simulation on the representative models; and (4) selecting, via the computer, the best representative model having a prediction value most similar to the observed data from among the representative models by comparing the observed data with results of the simulation,
(b) creating the final models and the soft data comprises (1) selecting, via the computer, the candidate models near the best representative model in a sequence of closeness from results of performing the distance-based clustering technique, (2) calculating, via the computer, the mean model of the selected candidate models, (3) determining, via the computer, whether the calculated mean model has converged, (4) if it is determined the mean model has not converged, setting, via the computer, the mean model as the soft data, and (5) if it is determined that the mean model has converged, setting, via the computer, the selected candidate models as the final models,
(c) the spatial correlation data and the hard data remain unchanged, and
(d) the hard data are either core data obtained from drilling or data obtained from geophysical well logging.