Document Type : Research Paper
Authors
Abstract
Well logs are principal sources of subsurface geological information. They provide significant information on mineralogical composition, texture, sedimentary structures and petrophysical properties such as porosity and permeability. By compiling data from various well logs, one can discriminate sedimentary units with comparable log characteristics. Sedimentary units with similar fluid flow and capacity are named rock type. Rock type determination is the most important task in reservoir characterization of oil bearing formations. Rock type may be determined using different data sets but their definition on the basis of wire line logs is most common. Multivariate cluster analysis (as the best method of data grouping) is one of the most accurate and effective methods in oil bearing reservoir zonation. The method is applied on both detrital and carbonates rocks. This method gets more support by improvements in algorithms and statistics. Proper combination of logs and appropriate algorithm will increase the accuracy, reliability and effect of the method. Similar faces may have different log responses due to diverse factors that affect the logs. Since using statistical methods and procedures are mandatory, in clustering procedure data are grouped with minimum distance and maximum homogeneity. It is obvious that distinct geological parameters can be related to a group of data, which are to be used by geologists for further interpretation. For this calculation, all log readings are considered as "observations" and the used logs as the "values of the observations".
There are several ways to compute the distance between objects. The "Standardized Euclidean" distance is used here in form the MATLAB software, because more accurate results are obtained with this procedure. By grouping log data in multidimensional space (equal dimensions with number of logs), each point (reading) can be related to a group of data (rock type). High resolution rock typing with reliable conclusions can be inferred with this procedure using pure mathematical formula in which there is no need to regression equations or trainings. In this method, any geological parameter described from other sources such as cores and thin sections can be related to wells with comparable rock types. The accuracy and reliability of defined rock types can be examined in wells from which suitable cores are available. Results from such a comparison provide a fundamental base for study of wells with poor core and cutting data.
Using MATLAB software, this study testifies a new simple method for rock type determination of Asmari Formation in Marun Field. The reliability of the method is examined by correlation of the resultant rock types with those of inferred from cores. Result from such a correlation indicates the reliability of method in rock type determination.
)