ABSTRACT
Importance of well-logging interpretation in reservoir characterization cannot be overestimated. However, it is not always possible to conduct a logging operation with a complete set of tools as some of them are not available at the field or economically expensive. Moreover, relationships used to evaluate petrophysical properties require prior knowledge of rock matrix properties which inevitably introduce a bias into the result of interpretation. The goal of this work is to build a general mathematical model based on the machine learning algorithms which enables to obtain porosity and permeability values accurately utilizing only well log data. The potential application of machine learning to this problem in comparison to the traditional well-logging interpretation is also explored. It is worth to note that the research is based on the actual information from several oil wells in Western Siberia.
