An example of neural network algorithm application to determine lithology using well logging data in a terrigenous section of the republic of tatarstan

Abstract

Accurate determination of lithology based on well logging data is an important task in the study of oil and gas fields. Large fields can include hundreds of wells, which affects the time required for interpretation. In this work, the authors tested the use of a neural network to determine lithology from well logging data. The geological structure of considered area includes rocks of the crystalline basement of the Archean-Early Proterozoic age and a sedimentary cover, represented by deposits of the Devonian, Carboniferous, Permian and Quaternary systems. Oil deposits are distinguished in the deposits of the Upper Devonian and Middle Carboniferous (Adbulmazitov R.G. et al, 2007). For the analysis, the authors selected a productive interval of Devonian terrigenous deposits. There are three learning models: "supervised", "unsupervised" (self-learning), and mixed (Darpa, 1998, Hertz et al, 1991). To solve the problem of finding the best porosity value set in this article, the supervised model was used. This means that during training, the neural network relayed on "correct answers" for input data. In the process of learning, the weights of connections between neurons in the network were adjusted in the way that the network gives responses closest to the correct result.