PNIPU: AI-based model for determining formation pressures

White Paper: Information Technology in the Oil and Gas Industry

2024: Model Development for Oil Production Control

Perm Polytechnic scientists have developed an artificial intelligence-based model to better control oil production. The university announced this on July 8, 2024.

In the oil fields, water is injected into the injection well to increase the oil recovery of the formation, which increases the pressure in the production well, thereby pushing the oil higher. To ensure that such waterflooding remains effective, it is essential to regularly monitor communication between them and ensure that water flows freely through the formation channels and into the correct location.

For July 2024, this is done with the help of expensive and long indicator studies. Scientists at the Perm Polytechnic University have developed an AI-based model that quickly and accurately determines the values ​ ​ of formation pressures depending on the volume of water injection. This approach will allow you to assess the quality of flooding of oil reservoirs with minimal labor costs.

Field development monitoring is an integral part of the overall oil and gas asset management system. It is carried out using various geophysical, hydrodynamic and special studies. Monitoring makes it possible to assess the energy state of reservoirs, control the dynamics of well saturation and much more, which ultimately affects the efficiency and quality of oil production. Now, thanks to information processing methods, these problems can be solved in more detail and more reliably.

This also applies to the assessment of the hydrodynamic relationship between injection and production wells. Water injected into the formation must flow freely between them to provide the necessary pressure to advance the oil. It is important to regularly assess the quality of this cross-country ability. For July 2024, this is done using indicator studies, when instead of water, a chemical reagent is injected into one well and its appearance is checked in another. But this process is very expensive and requires a prolonged shutdown of the entire production process.

Perm Polytechnic scientists have developed an approach by which you can determine the quality of communication between wells in a couple of minutes and without stopping the workflow. It is based on a comparative analysis of the average monthly formation pressure in the production zones and injection volumes of injection wells. The model based on artificial intelligence is implemented in the form of a specially developed software product. It allows you to reliably determine reservoir pressure even with a minimal set of input data.

The initial data are files that are uploaded from standard hydrodynamic models and contain information on the values ​ ​ of the average monthly well production rates (oil production volume) and the coefficient of its operation. The calculation duration is not more than one minute even for large development targets, and the calculation results in data on formation pressure values in the sampling zone of each well for each month of its operation. They are presented in the form of generalized and individual graphs, and are also unloaded in the form of a standard electronic table, - said Inna Ponomareva, Doctor of Technical Sciences, Professor of the Department of Oil and Gas Technologies at PNIPU.

The developers checked the program at the field with severe geological and physical conditions for oil production and established the complex nature of the interaction between injection and production wells. If the reservoir pressure in the producer does not respond to changes in water injection in the adjacent injection, this is an indirect confirmation of the lack of hydrodynamic connection between them. Comparison of the result with the conducted indicator studies confirmed the performance of the model and the feasibility of its use in practice.

The developed approach of Perm Polytechnic scientists has a fairly high predictive ability. On average, the reservoir pressure prediction error does not exceed 5%, which is a good result, especially in conditions of complex carbonate reservoirs. The program based on artificial intelligence solves the tasks of monitoring the development of oil fields with minimal labor costs and with a small amount of used geological and field information.