Reservoir characterization

In Search of Data-Driven Oilfield Reality

Many startup companies promote the wonders of analytics but few can ensure that the methods used and data generated live up to the hype.


A big problem facing big-data-driven petroleum engineering is convincing engineers that it is reality driven.

Birol Dindoruk,SPE Technical Director,Management and Information


It is an area of interest going back decades for Birol Dindoruk, who will focus on data-driven issues as he begins his term as the new SPE director for management and information.

The chief scientist for reservoir physics at Shell coauthored a 2004 technical paper (SPE 89030) about using pressure/volume/temperature (PVT) data to predict the properties of oil from wells covering a large area in the Gulf of Mexico. That was built on work from the late 1980s involving digitization of well logs and analysis charts.

The need for widely accepted standards that ensure that data and analysis accurately portray reservoir conditions is not new. But there are a lot of new puzzles to work on.

In 2005, Dindoruk wrote that, “We did not have certain concerns that we may have today because we did not have good-quality data and the detailed information available now.”

In 2018, new concerns have been spawned by the explosion in the number of sensors, the Internet, and far-faster data processing as the industry confronts challenges such as maximizing oil production from rock that was not even considered producible when that was article written. While there is a lot of discussion in the industry about analytics, Dindoruk said that based on what he has heard at SPE workshops, “approximately 90% of the work is getting the data in a digestible form” for analysis.

At the moment Dindoruk is working on projects on both the physical and digital sides of the problem. The physical side issues can be reduced to “is the data right?” Dindoruk is contributing to an SPE technical report on PVT reservoir fluids analysis. The goal is to meet the pressing needs of the industry, such as classification of fluids, with measurement methods delivering accurate, comparable, affordable data.

The question on the digital side is “is it real?” Can the output of analytical tools such as pattern-recognition algorithms pass muster with experts both in petroleum engineering and statistical methods?

The fluid properties report is a collaboration with Tom Blasingame, SPE’s reservoir technical director. It is an example of collaboration among technical disciplines to generate critical inputs for a variety of industry uses.

Fluid properties are as fundamental to reservoir engineering as salt and pepper are to cooking. They can be used to predict production, configure surface facilities to match the changing makeup of production, and predict what methods are likely to enhance production, among many other things.

The technical report, which is expected to be published during the second half of this year, will cover current best practices for collecting data and measurements to build large, statistically sound, databases. They need to be stored in a form that makes them readily accessible and compatible with other data, including older data sets. And there needs to be more “statistical tools to make sure there is not a bias in the data.”

As he begins his work as a technical director, Dindoruk is faced with setting priorities for further work that will support a range of disciplines. There are far more possible problems to consider than anyone has the time or money to solve. Dindoruk will be asking his advisory committee to help him rank what questions need to be addressed for SPE members

One question he will ask to narrow down the list of possibilities: “What if this was like Christmas day, what do you want out of the data? What are you missing?” he said. He will be looking for problems that can deliver answers that can be used in many ways, like PVT data, and so critical “you would be willing to pay for it.”