Using Statistical Techniques to Identify End-Members for Allocating Commingled Oil Samples Produced From Unconventional Reservoirs

Abstract

High resolution GC (HRGC) data can be used to determine the contribution from each of several zones to commingled oil samples. But allocation requires the availability of “end-member” samples from each zone. If core extracts are used to search for end-members in “shale” reservoirs, the number of potential end-members can be too large to evaluate using HRGC data. However, several statistical methods can be used to resolve this difficulty. A Hierarchical Cluster Analysis (HCA) assigns oil samples and core extracts to different groups based on the similarity of their composition. This method provides insights into the number of distinct zones in a shale reservoir, but HCA results cannot be easily used to identify end-members because they reduce all differences to a single variable (% Similarity) – a simplification that obscures underlying relationships. HRGC peak-height ratios form distinctive patterns on star diagrams that also can be used to identify different groups of oil samples and extracts. In addition, relationships between different kinds of patterns can be used qualitatively to identify potential end-members. Principal Component Analysis (PCA) of HRGC peak-height ratios is the most useful statistical method to identify end-members when isolated single-zone end-member oil samples have not been collected. The PCA method converts the values of peak-height ratios into new, independent (orthogonal) variables called principal components (PCs). The first PC explains the largest portion of the variance, and each succeeding PC explains additional, smaller parts of the variance. Produced oil samples and core extracts with similar compositions have similar PCA case scores. That relationship persists on three-dimensional case-score figures created using PC1, PC2, and PC3 values, and on each of the three analogous two-dimensional figures (PC1 vs. PC2, PC1 vs. PC3, and PC2 vs. PC3). Mixing lines between end-member samples can be readily identified using 3D and 2D PCA case-score figures. A putative mixing line initially identified using only one 2D case-score figure (e.g., PC1 vs. PC2) can be evaluated rapidly by determining if the same line exists on other 2D figures. PCA variable-loading figures can be used to determine how specific HRGC peak-height ratios influence the locations of oil samples and extracts on PCA case-score figures. This paper illustrates these principles using HRGC data measured on oil samples and core extracts from several shale reservoirs.

AAPG Datapages/Search and Discovery Article #90323 ©2018 AAPG Annual Convention and Exhibition, Salt Lake City, Utah, May 20-23, 2018