Central AsianResearch Index

Automating conventional well correlation using machine learning techniques

The fundamental aspect of well correlation involves identifying the sequence of rock stratification. This correlation serves as the basis for constructing a geological model that accurately represents the structure of oil and gas deposits. Consequently, the precise estimation of reserves, reliable oil recovery factor, and valid field development plan largely rely on the correct implementation of this correlation. The conventional approach to well correlation is a time-consuming, particularly when dealing with a growing number of wells. In this study, we present an automated model that replicates the conventional well correlation process by employing the Minimum Spanning Tree (MST) and Dynamic Time Warping (DTW) algorithms. Our proposed model initially pairs all wells based on the shortest distance using the MST algorithm. Subsequently, the wells within each pair are correlated using the DTW algorithm. The implementation of our proposed model on the wells of the Karatobe Oilfield in Kazakhstan demonstrates its reliability. Moreover, this model can be readily applied to other oil and gas fields, offering a highly effective alternative to the traditional, time-consuming methods of well correlation.

Сведения

Журнал
Neft i gaz
Опубликовано
2024 · Т. 143, № 5, С. 30-42
Язык
Английский
ISSN
1562-2932 · 2708-0080
Издатель
Oil Research and Engineering Center of the National Engineering Academy of the Republic of Kazakhstan