The Gazprom Neft Science and Technology Centre and the Skolkovo Institute of Science and Technology (Skoltech) are developing a unique digital technology for the selection of optimum oilfield development systems. This bespoke digital tool will make it possible to reduce the use of commercially competitive products, saving the company approximately RUB200 million by 2022.
Innovative algorithms, developed as part of the “In-flux Meta-model” project* will allow exponential reductions in the time taken in determining the optimum number of production and injection wells to be drilled, taking geological preconditions into account, as well as allowing production to be monitored once they go into production. Machine learning technologies (including deep neural networks) are being used in developing the software. The project is being implemented as part of the company’s Technology Strategy, under its “Electronic Asset Development (EAD)” initiative.** Building a high-quality hydrodynamic model of oil reserves is essential in field development. Questions as to how many wells are needed, of what type, and where they might be best placed, within the reservoir, also have to be answered. The optimum choice for each field is determined individually. Commissioned exclusively for the Gazprom Neft Science and Technology Centre, Skoltech employees are developing a meta-model that performs the necessary calculations on the Institute’s supercomputer, and analyses the results obtained.
Mars Khasanov, Director of the Gazprom Neft Science and Technology Centre, commented: “Creating new digital products that optimise the company’s technological processes is fundamental to our development. Imagine an individual with extensive experience in solving thousands of similar equations, each with different parameters. His experience is so extensive that he can predict — very accurately — the answers to new equations. This is the same thing — we have learned from the equations that we know, working them out hundreds of thousands of times, and we can predict the importance and impact of metrics and parameters of interest to us, on the basis of new data.”
* A meta-model is a model built on machine learning techniques. Such models “teach themselves”, processing large volumes of data and taking previously identified patterns and trends into account in subsequent operations.
** “Electronic Asset Development (EAD)” is Gazprom Neft’s strategy for developing digital initiatives in all areas of exploration and production, including geological prospecting, geology, drilling, development, production, and field construction. Under implementation since 2012, the EAD initiative became part of the company’s Technology Strategy in 2014, and is now a key area of focus. The company’s technology strategy was updated in 2017 in response to new production and business objectives of various Gazprom Neft subsidiaries. More than 30 projects are currently in hand under the EAD initiative.