ISSN 1991-2927
 

ACP № 3 (65) 2021

Author: "Dmitrii Vladimirovich Kurganov"

Dmitrii Vladimirovich Kurganov, Candidate of Sciences in Physics and Mathematics; graduated from Samara University; Associate Professor at the Department of Oil and Gas Engineering of Samara State Technical University; an author of articles, patents in the field of the modeling of oil and gas field development processes. e-mail: Dmitri.Kourganov@inbox.ruD.V. Kurganov

The calculation of the bottomhole treatment success probability using machine learning techniques59_6.pdf

Machine learning is the most widely used branch of science and engineering nowadays. The availability of electronic real- wide information is an important condition for implementing the machine learning. Through the long history of exploitation of oil fields, a significant database related with the wells’ development and applied techniques stimulating production was created and accumulated. The article deals with one of the machine-learning methods of analyzing the prediction of geological and engineering operations implemented in the producing oil wells. In particular, by the example of database data including the results of hydrochloric-acid-and-mud-acid jobs implemented in the oil fields of Ural-Volga region, as well as in terms of specific decision-tree models, the probability of successful geological and engineering operations is calculated, and recommendations on the selection of factors allowing to optimize these operations are given. The openness, average number of permeable intervals, reservoir temperature, actual well watering, fluid properties are parameters influencing the geological and engineering operations efficiency. In many cases, it is difficult to predict the influence in the future especially if other factors are available. The existing analytical models could not describe fully the factor variety in processes running in a bottomhole zone, especially in the context of nonlinear flotations, physical and chemical interaction between formation fluids and injection solutions. The described technique allows using any number of key factors and any combination of them, as well as detecting the most important of them including the parameters described, but not limited to. In this case, the application of decision tree models is an intuitive way allowing to select neatly sampling attributes at each level by the use of algorithms. The article also describes in detail the algorithm of sampling attribute calculation. Decision tree methodology can be used for solving other problems in the oil producing industry with significant practical experience.

Hydrochloric acid treatment, mud acid treatment, bottomhole treatment, yield, crude oil, oil well, probability, machine learning, decision tree, modeling.

2020_ 1

Sections: Information systems

Subjects: Information systems.



© FRPC JSC 'RPA 'Mars', 2009-2021 The web-site runs on Joomla!