Computational tool lowers cost, improves exploration accuracy

By Patrick C. Miller | November 19, 2014

A new computational tool developed at Oak Ridge National Laboratory can take some of the cost and guesswork out of oil and gas exploration.

Clayton Webster, Oak Ridge senior scientist and leader of the computational and applied mathematics group, said the new approach uses a multilevel Monte Carlo method involving computational algorithms. Repeated random sampling with the tool provides a more accurate picture of the earth’s subsurface. 

“What we wanted to do is reduce the overall uncertainty or the confidence in where drills are placed and where we get the biggest bang for our buck,” Webster said.

According to Webster, the computational tool developed by the Oak Ridge research team is a perfect fit for oil and gas exploration.

“We develop a lot of techniques like this, but it’s rare that we will go out and directly apply them to an application,” he explained. “Typically, we’ll develop a method independent of an application and let others apply it. But in this case, we decided to apply it directly because it was a very good sales pitch for what we we’re doing.”

The approach is especially effective because the heterogeneous nature of the earth's surface causes a large number of model parameters.

“At the end of the day, what we’re really interested in showing is that we reduce the overall computational cost associated with the procedure,” Webster said. “As long as the procedure is ensemble-based and can be written in a hierarchical fashion, you can use what we did and it can be generalized to all these different kinds of problems using the data directly.”

The Oak Ridge computational tool can result in a dramatic reduction in computational costs, Webster noted.

“It’s somewhat problem dependent, but I can tell you from the kind of problems we looked at—simple exploration type problems—that we were able to see cost reductions on the order of up to 50 to 60 percent,” he said.

“It means you can go further and you can do more exploration,” he continued. “If you’re able to afford some amount cost in the exploration process, then you can go a lot further and get more confidence in your results.”

Perhaps the best part is that anyone involved in oil and gas exploration can apply the tool to their computational method.

“You can take the approach that we’ve done and any exploration company or individual looking at these applications can essentially do exactly what we’re doing by using their existing technology, but then adding on this ability to balance errors as you go through some type of hierarchal sampling or optimization procedure,” Webster said.

Because it’s becoming more difficult to develop new methodologies for exploration optimization, the type of math research conducted at Oak Ridge is a rapidly growing field, Webster said.

“Instead of just doing your plain vanilla sampling procedure for optimization, if you add in this ability to balance errors in this multilevel framework, then you can still keep your same approach, but have dramatically less cost to do it,” he said. “That’s why we call it smarter sampling for exploration because we’ve come up with a way to dramatically improve the procedure that people have been using for a long time.”

Webster and his colleagues D. Lu, G. Zhang and C. Barbier will soon be publishing a paper on their research entitled "A multilevel Monte Carlo approach for predicting the uncertainty in oil reservoir simulations. He said the paper can be used by those interested in applying their approach. In six to eight months, Oak Ridge should have a software application available.