Open Collaborative Research
Applied mathematicians from IBM Research, such as Andrew Conn, are working with the Norwegian University of Science and Technology to maximize oil exploration in the North Sea.
Along with commodities such as copper, corn and cattle, oil has the power to shape the quality of our daily lives. No assertion is more self-evident, and yet nearly everything associated with oil -- especially finding it and getting it out of the ground -- heads the list of challenges facing every developed and developing nation today. For that reason, applied mathematicians at IBM Research are looking for ways to help oil companies find oil faster, quicker and more cheaply than their competitors.
One such project -- "Reservoir Management and Production Optimization" -- seeks to develop algorithms -- used in mixed integer nonlinear programming problems – to optimize petroleum production network simulator parameters using proxy models and structural constraints. To expedite the enormously complex process of managing, improving and controlling underwater exploration and production, the project, part of IBM Open Collaborative Research (OCR), will also make open code available to developers interested in using or improving it.
How the project got started
Andrew Conn (pictured), an advanced analytics and optimization scientist, was working with Statoil, a Norwegian energy company, when the Center for Integrated Operations at the Norwegian University of Science and Technology asked him to sit on its technical board. The appointment in 2006 meant a trip to Norway for a couple of days every six months -- a lot of travel for such a short visit. To make the trip worth his while -- and IBM's -- Conn got more involved with the Trondheim-based university, working with NTNU summer interns at IBM Research every year and initiating the OCR project that would focus on optimizing the extraction of oil and gas from the sea. The goal: Create models and simulate numerous scenarios to locate and manage petroleum in the North Sea more rapidly and efficiently.
Now in its second year, Conn's OCR project continues to develop simulations and models that will help energy companies maximize the amount of oil they can get out of a reservoir basin. And as with any underwater exploration, the search is always complicated by the difficulty of getting a clear picture of the tremendous geological diversity in the reservoir basin, including rock fractures and seismic activity.
How the team is going about it
The OCR team has relied on various optimization techniques. Two of the most popular ones: Line search and trust region.
Using the line search method, researchers determine where they will begin their exploration and in what direction they will go. Using the trust region method, researchers create a model, which they compare to the actual region under exploration. If there is reasonable agreement between the model and the behavior of the actual region, then the researchers will expand their model to include a wider area of potential exploration. Conversely, if the actual region does not behave like the model, the researchers "downsize" their model. The goal here is to home in on a small region and create as precise a model as possible of it.
The team has also worked at simulating the entire field that they are trying to optimize. The model for these simulations include the various components -- wells, pipelines, a manifold and a separator -- that make up an exploration application. Simulations might be done for pressure drops on the pipelines or for how wells behave, to name two instances. As part of the project, Conn and his colleagues at the Center for Integrated Operations compared an IBM model with NOMAD Black Box optimization software often used in optimization without derivatives for complex simulations. Where Conn's algorithm required four iterations, NOMAD's required 351. Where Conn's team needed to render 82 well simulations, NOMAD needed 23,402. Likewise, Conn's team rendered 1,662 pipeline simulations; NOMAD's needed 15,602. And the IBM-NTU simulation had a greater than 10 percent improvement over NOMAD’s approximate optimal solution.
In addition to designing algorithms to optimize underwater petroleum extraction, the project is also positioning IBM as the go-to subject matter expert.
“You would be surprised at the number of people who don’t know that IBM is engaged in this kind of work,” Conn says. “They can’t believe that we are in the business of helping oil companies save money by optimizing their exploration processes. Through the OCR, we are getting people inside and outside the petroleum industry to understand that if you can improve the reservoir models by even one percent, you are going to save companies millions and millions of dollars.”
Last updated on April 30, 2012