Development of the Approach for Automatic Well Logging Interpretation for Big Number of Wells with the Use of Machine Learning

Summary The increased interest for geological and reservoir simulation model construction for the old fields raises the issue of reinterpretation of well logging data of the old well stock, taking into account the concepts of geology, specified during the period of field production. This work shows the results of mathematical approach development for automatic interpretation of well logging data of the old well stock. The work is aimed to solve the problem of fast reinterpretation of a standard logging set using machine learning algorithms. The solutions obtained for the determination of stratigraphic boundaries with the use of logistic regression and the detection of lithotypes basing on the support vector machine are presented. Mathematical algorithms and approaches to use them are presented and described.