Release:2019, Vol. 5. №1
About the authors:Igor A. Muravev, Senior Lecturer, Department of Software, University of Tyumen; email@example.com
Identifying the properties of oil and gas reservoirs based on information obtained from well logging is one of the main areas of research in the field of geological and hydrodynamic modeling of the reservoir. The insufficient effectiveness of accurate mathematical models for analyzing well survey data, as well as the large volume and noise of these data, determines the relevance of using machine learning methods to identify reservoir features.
This article investigates the possibility of classification of terrigenous collectors using various methods, including support vector machine, decision tree, gradient boost, random forest, and multilayered neural network. The data set was formed on the basis of well logging curves for 24 wells of one reservoir. For training classification models, pre-normalized data from inductive logging, lateral log, neutron-neutron logging on thermal neutrons, borehole electrical measurements, resistivity logging, spontaneous potential logging, gamma logging, and resistance logging were used with five different gradient sondes. To assess the accuracy of classification models constructed using various methods, in each case, cross-validation was performed, the average value of accuracy and standard deviation were estimated. For the support vector method, the influence of the choice of core function (linear, polynomial, and sigmoid) was investigated. In the case of a neural network, its architecture varied, including the number of hidden layers and neurons, activation functions on different layers, and the probability of a dropout. The quality of the obtained classification models was also evaluated by the values of the elements of the confusion matrix.
The results of computational experiments have shown the effectiveness of the use of machine learning methods and, in particular, multilayer neural networks to identify with high accuracy (about 90%) of reservoirs with oil.