Conventional methods of statistical analysis often break down in the study of reservoirs because exclusionary processes reduce the number of elements studied until they are statistically meaningless.

Using an Enhanced Oil Recovery (EOR) Field Test Data Base which was developed for 187 field tests, methods of multivariate analysis, particularly cluster techniques, were used to look for similarities and differences in the data. Each field test was treated as a key element with the following nine reservoir parameters considered as independent variables: porosity, permeability, oil saturation, API gravity, initial water saturation, age, depth, net pay, and viscosity.

A 187 × 12 unweighted data matrix was constructed. Then a 187 × 187 diagonal matrix of similarity coefficients was calculated using a moment based equation. The similarity matrix was ordered and plotted in the form of a dendrogram using a pair-wise grouping technique.

Clustering effects were found correlated to the five different enhanced oil recovery processes used in the field tests. The processes involved are in situ-combustion, carbon dioxide injection, improved waterflood, surfactant-polymer injection, and steam flooding.

The application of these methods to critical parameter analysis of a field test data base for enhanced oil recovery are discussed and illustrated by an assortment of computer and display techniques. The methodology appears to have significant potential in evaluations involving selection and application of reservoir screening criteria, the identification of minimum data requirements for decision making, audit methods for the examination of data bases, and comparative analysis of large numbers of reservoirs simultaneously.

An exploratory approach to prediction of performance of EOR Field Tests using an interactive stochastic model will also be described.

End_of_Article - Last_Page 546------------