The final of the Gazprom NeftSmartOil Contest — the first ever competition investigating ways of using machine-learning technologies to improve efficiency in oil production — took place on 13 October, organised by the Gazprom Neft Science and Technology Centre, together with Innopraktika. The key objective of the competition is to find new solutions in data analysis. Winners, together with employees from the Gazprom Neft Science and Technology Centre, could go on to form part of a team working on developing a digital “AI assistant”, whose key function will be processing large volumes of incoming field data in order to determine optimum technological and economic solutions in oil production.
The outlook for the development of IT technologies in the oil industry is examined in this interview with Mars Khasanov, CEO of the Gazprom Neft Science and Technology Centre.
— Oil was, until recently, one of the last industries one would associate with IT technologies: but now artificial intelligence (AI) technologies are making their way into various economic sectors — and significantly disrupting them. Why, in your view, is the use of new scientific approaches now necessary, even in the oil world?
— Actually, the oil industry has always been closely associated with IT technologies. Oil-bearing strata, after all, is located some 2.5 below ground, and its properties can only be assessed in terms of indirect measurements. So it’s always been necessary to analyse information obtained in developing an oilfield — to simulate or model it — and undertake mathematical experiments. It’s no accident that the oil industry was one of the first to keep tabs on the question of digitisation — we have always worked with information technologies, and have been following all changes in this area closely.
— Аnd have you already found any particular benefits from using new technologies?
— Of course. Nowadays, as the high-quality reserves developed during Soviet times are coming to an end, we are drilling at blocks characterised by a porosity of 0.1 millidarcys. Porosity reflects a strata’s propensity for allowing oil to pass through it. And these days it is, at a minimum, a thousand times lower than at fields drilled in Tatarstan or Eastern Siberia during the Soviet era. In developing these low-permeability reservoirs — what are known as “hard-to-recover” reserves — we use horizontal wells: these are high-tech facilities, and it’s very important that they are properly designed. In order to understand how tightly together wells need to be placed, how long the horizontal shaft needs to be, or how much fracking operations need to be done, complex non-linear mathematical calculations have to be solved, and a multitude of iterative calculations have to be made. And you can’t do that by hand. Every field involves a complex system of strata, wells, and surface and infrastructure facilities. There are so many criteria here driving efficiency that finding a global extremum is only possible with the help of artificial intelligence.
— But you also need people trained in working with “Big Data”, surely? Are there enough of these specialists?
— If we’re talking about Russia as a whole, then, fortunately, there are. All oil companies need people able to translate a technological challenge into the language of mathematics. Gazprom Neft is in a good place as regards physicists, mathematicians, and IT specialists. We have always paid attention to the scientific element at our Science and Technology Centre — developing scientific oil engineering, so we were ready for the challenges of the digital epoch.
— According to data from PriceWaterhouseCoopers, 74 percent or Russian companies plan to invest in AI in the next three years. What policy is Gazprom Neft pursuing here?
— We’re developing our own software products as one of the areas of our Technology Strategy — that is, our ERA initiative (abbreviated from the Russian "электронная разработка активов) — Electronic Asset Development. We started doing this in 2012, before anyone was talking about digital transformation. We began developing our own software products because we understood it was necessary to move away from imports here. At the same time, we’re not just working around import substitution but, I would say — “import out-gaming”. Many of our software solutions contain mathematical ideas, algorithms and applications that foreign alternatives don’t. Added to which, we already have, within the company, tools based on AI technologies — such as, for example, “ERA Optima”. This is a robot programme that uses mathematical optimisation algorithms to find the best solution for field development. The best, moreover, not just in terms of strata geology, but also in terms of economics. Every field has its own infrastructural and technological limitations, and taking all of these into consideration, in order to increase oil production without incurring extra expenditure, is sometimes difficult for someone to do. Whereas a robot takes all this into account and helps optimise our decisions. We have other solutions, too, for various areas. For example — machine learning in drilling: AI can predict any overrunning of productive strata, so that we can correct the trajectory successfully. There are tools for the very earliest stages — for example in looking for similar fields: AI can help evaluate fields that have not been fully investigated, comparing these to similar, already drilled fields.
— The tools you’re talking about — are they designed to be used by various specialists — that is to say, individually? Or is work undertaken holistically?
— Yes — ERA is a complete, integral strategy, an ecosystem of software products. At the exploration stage a field is thrown into a cohesive information space, and thereafter, at every stage — from development to oil and gas production — we work with its digital twin. We analyse information from production wells and use this in order to improve the model, to obtain more accurate forecasts, and to understand where there is remaining oil in place (ROIP) — while, at the same time, monitoring OPEX. That is — we are modelling, designing and overseeing the construction and utilisation of all oil facilities, from start to finish.
It’s important to understand that machine learning, despite its different applications, is generally used to address only one kind of task. Cognitive technologies help us analyse unstructured information. The volume of information available to engineers for analysis is growing every year, but only 10 of this is presented in a standard, structured form — like tables, structured databases, and so on. Ninety percent of information is stored in an unstructured form. There isn’t enough time or people to get this set out in front of you. In some areas AI, when used in working with big data, can save up to 80% of the labour involved in routine operations, according to data from our “Cognitive Geologist” project. As part of this project we have developed a programme which allows the most useful information to be extracted from existing data, thereby speeding up the entire process of geological exploration.
— Do you have any joint partnership projects in this area?
— Yes, we’re working with a large number of scientific centres and universities in Russia. Apart from which, we are collaborating — developing software products and joint technologies — with companies including IBM, Yandex, and Skoltech, as well as other companies in the digital arena. We’re proactive in drawing out ideas, in order to develop our own software products.
—Are there similar strategies to ERA in place at other companies?
— Separate, individual parts of such systems — yes, of course there are. But we’re the only ones to have adopted a systematic approach like this, and to have involved such a wide range of tools. Few oil companies are programming independently to this level.
— You’ve talked about a number of promising projects that are helping extract useful information from large volumes of data. Can you give us any examples?
—Actually, we find patterns and trends in bulk data every day, gaining new knowledge and insights from this. About half a terabyte of data is generated every day in the Science and Technology Centre alone. Obviously, this is all processed, with our specialists drawing important conclusions from it.
But we do have a project on which we’re processing a massive volume of historical data — the “Lost Horizons” project. When the Western Siberian fields were being drilled out in the 1960s and 70s, those strata that were known to be oil-bearing were studied in depth. Those strata for which geophysical data was ambiguous were often overlooked. And now, when the thick seams are finished, life itself is forcing us to use new technologies on low-permeability strata. So we thought: why not go back and re-process the entire mass of data we’ve accumulated? As a result, we’ve been able to compile a map showing the distribution of thin-bed seams. And high-tech wells mean we can produce enough oil from these to make it financially viable.
— So — turning to international experience. What is current practice in using AI throughout the oil industry, worldwide?
— There is one thing that marks our company out. Practically all oil companies are applying achievements in digital technologies, in AI in particular, in operating oil fields. Take a typical task: a pump is in operation at a well, with receivers attached to it measuring its vibrations and — based on how it oscillates, predicts when it might go wrong. That sort of predictive analytics is widely developed, both in Russia and abroad.
What are we doing? We’re looking a bit wider. The real value in using technological solutions becomes apparent on launching a project, when we’re involved in exploration and are doing the conceptual development. Once we’re into operation we can also optimise things, but this only increases efficiency by 10 to 15 percent. Whereas at the concept stage we can optimise things by 50, 70, 90 percent, reduce costs two-fold, and increase production. Indeed, it’s at precisely this stage that we determine how frequently we need to drill wells, what kind they need to be, and how many clusters — the areas in which wells need to be drilled — are required. These decisions impact financial expenditure, so our main focus is on this, on making the right technological decisions.
—And Gazprom Neft is in line with global trends here?
— I would say we’re ahead of the curve. You don’t need to study western experience in order to be one of the leaders of the digital revolution — in that situation, you’re no longer a leader, but a follower. To be a leader you need to make your own way, find your own — your best — route to follow. Gazprom Neft is taking precisely that strategy. We’ve correctly identified the place where we need to be.
— How is the oil industry going to look in 20 to 30 years’ time, thanks to AI, and thanks to high technology?
— I think we’re going to be able to maintain oil and gas production at the same level, notwithstanding the fact that oil reserves are going to be harder and harder to produce. The productivity of the fields we’re drilling over are plummeting, but information technologies are making it possible for us to cut costs, and work out technological solutions correctly. Coupled with new materials, this will allow us to produce oil from declining reserves, without increasing production costs.