PERFORMANCE PREDICTION FOR DEEPWATER GULF OF MEXICO USING DATA MINING

Abstract

The estimation of recovery factor is important in every stage of hydrocarbon development, and multiple traditional techniques are available for engineering application for particular fields. However, the estimated recovery factor can be very different using these methods. This is particular true for deepwater development as the parameters associated with the recovery factor estimation have significant uncertainties. The objective of this study is to apply the data mining technologies based on the data from the developed fields in Gulf of Mexico. Using database of 395 Deepwater Gulf of Mexico (dGOM) oilfields with 84 attributes, set of dimensionless variables are calculated; and these dimensionless variables are used as the input for data mining with an aim of obtaining the recovery factors. A subset of 59 oilfields that have water drive mechanism are selected for discovering the generalized correlation for recovery factor using data mining techniques. In the study, a variety of data mining techniques such as K-means and principal component analysis are used for classifying oilfields into four categories. Subsequently, partial least square (PLS) regression is used to relate the dimensionless variables to the recovery factor from sparse data in dGOM. However, not all clusters show very high coefficient of correlation, hence limiting the applicability of this method. This study shows that dimensionless numbers, together with data mining techniques, can be very useful to predict field behavior in terms of recovery factor for sparse datasets with widely scattered reservoir properties. This information can be further used for the preparation of data acquisition and risk assessment plans to set up a framework for decision-making on risks and uncertainty for optimizing reservoir management and production forecast.