Researchers Use Machine Learning to Enhance Oil Recovery

In a new development, a team of Skoltech scientists and their industry colleagues have discovered a way to use machine learning to gauge thermal conductivity of rocks. This is a crucial parameter to enhance oil recovery, especially from shale.

Rock thermal conductivity is crucial for modeling a petroleum basin as well as to design enhanced oil recovery methods. Meanwhile, tertiary recovery methods allow oil field operators to extract significantly more crude oil than what is extracted using basic methods.

Among many, thermal injection is a common enhanced oil recovery method. Using this, oil in the rock cavity is heated employing various means such as steam, and to use the method, it requires extensive knowledge of heat transfer procedures within a reservoir.

Lack of Satisfactory Results of currently used Tertiary Methods led to use of Indirect Ones

The use of any of these processes would require in situ measure of rock thermal conductivity. Nonetheless, this has turned out to be a herculean task that has not yet produced satisfactory results for practical use. To address this, scientists and practitioners resorted to indirect methods. These methods help to infer rock thermal conductivity using well-logging data. This is inferred using high-resolution picture that are generated of vertical variations in physical properties of rocks.

At present, three core problems do not allow direct measure of thermal conductivity within non-coring intervals. Firstly, it is time required for measurement. This is because petroleum engineers do not allow to put the well on hold for a long time, as it is economically not viable. Secondly, convection of drilling fluid that has been induced has a drastic impact on result of measurements. Finally, unstable shape of boreholes, which has some association with technical aspects of measurements.