Development of approaches to automated correlation from well log data using machine learning

Latypov I.D. “RN-BashNIPIneft” LLC (“Rosneft” PJSC Group Company) Ufa
Markov A.V. markovav@bnipi.rosneft.ru “RN-BashNIPIneft” LLC (“Rosneft” PJSC Group Company) Ufa
Evgrafov N.A. “RN-BashNIPIneft” LLC (“Rosneft” PJSC Group Company) Ufa
Shagimardanova L.R. “RN-BashNIPIneft” LLC (“Rosneft” PJSC Group Company) Ufa
DOI: 10.24412/2076-6785-2024-4-47-51

Abstract
This paper discusses the principles and methods of in-situ section correlation and examples of its application to enhance the quality of petrophysical interpretation. One issue with automatic in-situ correlation is its dependence on the order in which wells are considered. To eliminate this problem, one option is to define the bypass paths of wells based on their proximity according to a Euclidean norm using log curve data. The paper presents an approach to automatic well log correlation using principal cluster analysis, component analysis and dynamic time warping.

Materials and methods
The paper discusses methods for intra-situ correlation of well sections.
Automatic section correlation is based on the use of cluster analysis algorithms, principal component analysis and dynamic transformation of the time scale. Principal component methods and cluster1 analysis are used to organize data from different wells according to geophysical responses, which allows for more efficient cross-section correlation using the dynamic time transformation (DTW) algorithm.

Keywords
well log correlation, k-means method, dynamic time warping, principal component method
Download article