We propose a novel approach to facies classification based on a supervised machine learning algorithm. This approach allows for the automatic facies classification on a field scale based on an ensemble of Decision Trees algorithm associated with gradient boosting. Major steps of the workflow include data integrity assessment, data scaling, identification and correction of gaps in data, log processing, feature engineering, training, testing, and tuning the hyperparameters on the validated set of data. At the ultimate stage of the workflow, the algorithm accepts a set of well logs as an input and produces a discrete facies type as an output. This method substantially increases the quality of the facies classification, that is key to further geological modelling and dynamic simulation that help reduce drastically the risk of incorrect well planning, fracturing and other operations, thus avoiding a huge negative financial impact. The novelty of approach is related to the selection of machine learning algorithms that are best fitting the dataset, combined with a workflow to enhance the dataset itself.
The paper describes the stages of developing of an integrated model for the oilfield in Western Siberia. The reason to create the model is the re-engineering of the field infrastructure because of the start of the program for drilling of several new pads. Drilling program is done together with the partial disassembly of the existing infrastructure of the old high water cut wells. Some specific features of work with initial data are presented in the paper. Some new approaches related to software customization are also presented. The organization of the process of performing calculations is interesting because the software was provided by different vendors, which is typical for almost all companies. The geological and hydrodynamic model was designed by using tNavigator, and the well and surface infrastructure models were designed in Weatherford software complexes - WellFlo and ReO. Integration was carried out by a specially developed software that carried out the transfer of boundary conditions between applications and saved intermediate results in the MS Access database. To increase the calculation speed and perform a series of calculations for different reengineering scenarios, the integration module is configured to use distributed computing on a supercomputer cluster. The number of software licenses allows user to flexibly use parallel computing and reduce the time to find a joint solution for a partially implicit scheme, bringing the time costs to an explicit calculation scheme. To process the results and prepare reports, standard office applications that are compatible with the database were used, which made it easier to prepare the final results. The results of the project show that in the late stages of field operation, key factor in making decisions on investments in the modification of the infrastructure systems (the choice between local injection with the use of local separators and the gas utilization system at gas piston power plants and centralized water pumping in the water injection system) is the ratio of costs for the construction of site facilities and linear objects. If there are prepared sites for placing separators equipment, and the distances from previously built central water pumping systems are significant, reconstruction of obsolete water lines is not profitable. Also, when making decision to develop local injection system, it is important to take into account if there are significant differences between injection wells, the main of which is injectivity coefficient and, as a consequence, the necessary wellhead pressure. Performing calculations on an integrated model allows the user to do comprehensive comparison of all scenarios for the development of the field infrastructure. The entire set of calculations is highly automated and upon completion of the calculations, engineers receive data for each production, injection well, the dynamics of reservoir pressures. Individual operation modes of ESP wells and energy consumption of the production system and water injection system are calculated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.