Data mining/analysis

Time To Enlist in the Analytics Army

Some operating companies are now enlisting engineers as foot soldiers in their analytics army. It is not required yet, but those looking to get ahead would be wise to get involved.

jpt-2018-06-dataanalytics.jpg
Source: Getty Images.

Data analytics is the future of ­getting ahead for engineers and geo­scientists in the exploration and production (E&P) business.

Three executives of large companies recently played up the growing importance of data-driven employee development, from data science boot camps at BP to the Citizen Data Scientist program at ConocoPhillips.

“We have got a data science community that is growing by leaps and bounds,” said Michael Rowley, director of technology and innovation at BP, during a presentation at the recent Data Science Convention 2018 put on by the Data Analytics Study Group of the SPE Gulf Coast Section.

“The most powerful way for us to promote data analytics is to give our technical people training so they are fully armed with data analytics tools,” said Greg Leveille, chief technology officer of ConocoPhillips.

Those willing to learn and apply these new tools “will have a disproportional impact on each of your companies,” Leveille said, adding “our industry will be populated by people who know and who can do data analytics well.”

In a tough job market, data skills could offer a critical edge. If two people are interviewing for a job in unconventional exploration, and one has analytics skills and experience and the other does not, “guess who gets the job,” said Andy ­Flowers, director of advanced analytics for Marathon Oil.

One sign of the times was the number of data scientists asking questions at the gathering. Another was the 400-person crowd at the first-ever gathering. They are building and maintaining the industry’s new data gathering and analysis infrastructure, putting large databases and analytical tools in the hands of engineers, geoscientists, and financial managers.

They are a small part of the E&P workforce, are hard to find, and costly to hire.

“We feel like we won the lottery when we get a new person in” for the data ­science team, Flowers said.

But engineers and geoscientists will remain central to E&P companies because “you cannot outsource knowledge of the oil industry,” Flowers said. They will be asked to use their experience to focus the work on the most important problems, and apply methods drawing on both ­traditional and statistical approaches.

“People with a good grounding and understanding of physics, combined with analytics, can solve problems that are very hard to solve” with first principles physics, Leveille said.

Increasingly, the data-driven tools will also integrate the economic aspects of decision making. Engineering decisions need to be aligned with current corporate goals. An application integrating technical and economic information could help an engineer tailor a gas project development plan based on predicted demand.

“Not everything is about geology and physics,” Rowley said. Another priority is to figure out “how to connect up data in the financial space,” he said.

Digital Imperatives

This is the latest digital imperative engineers have faced since computers began changing the face of the oil business years ago.

“It is a journey much like the journey we had installing computers,” said Leveille, recalling the days when using a computer required signing up for a time in the room that housed them.

Troy Ruth’s story about becoming a data scientist began at an isolated oil industry outpost in Indonesia where he occupied himself as a child by learning to program a personal computer. His father, a geophysicist, was working on one of the first workstations.

Now Ruth is CEO and chief data scientist for Ruth’s Analytics & Innovation, whose key product is designed to create an easily manipulated database program (based on NoSQL) designed to provide quick access to multiple types of data tools for easy manipulation.

He was one of several consultants on the conference agenda selling ideas designed to expand the analytics user base beyond early adopters willing and able to use the available tools to cobble together their own solutions.

Corporate training programs can be divided into those seeking to be power users—workers seeking a deep understanding of tools such as Python, a widely used programming language for data analysis—and those who want to be proficient day-to-day users of the tools needed to search and process data.

The tools are not the hard part of this transition. For traditionalists, the challenge is accepting that some aspects of petroleum engineering are better understood statistically using regression analysis than by the physics-based petroleum engineering they learned in college.

“The marriage of subsurface and data science is a tough one,” Flowers said. It is tough because there is a division between those committed to traditional methods and those who believe statistically driven analysis is the only way. Those leading this transition are cast in the role of marriage counselors trying to persuade battling partners to accept that each of them has a valid point of view.

The future will be a mix of physics and statistical analysis. Unconventional E&P forced the issue because the behavior of wells in ultratight rock often could not be modeled using classical models. But traditional thinking and experience is needed to target the problems likely to yield the biggest payoff, and reject results which are not physically possible.

“Anyone who thinks they can stay within my old suite of technologies, or dump all my old technologies and use only analytics; both of those will not turn out well,” Leveille said.

Required Classes?

Those who are talking up the importance of advanced data skills were raising questions about how best to educate workers. Flowers and others called for colleges to incorporate basic training into programming language such as Cobra.

At the University of Texas at Austin, petroleum engineering students can take courses on analytics, said Pradeep Ashok, a senior research scientist working on drilling automation who sees many students embracing data-driven methods.

A sign of the student interest in using data is the number of students signing up for classes where they spend a semester analyzing actual drilling data sets. Based on their insights from these big, messy datasets, they offer advice to the companies that donated it.

UT is collecting a variety of E&P data to prepare students to compete in a data-centric world. Ashok said training in school and at work is needed, but there is a limit to what many users need to know.

“You just need a few techniques,” he said, adding, “You could have a full program on data science but I do not think it is necessary.”

Employers also see the interest. Flowers said “some of our employees are naturally gravitating” to these new ways of thinking. “The question of whether college students are getting enough exposure to new data-driven methods will be discussed at the 2018 SPE Petroleum Engineering Education Colloquium this August in Katy, Texas, just west of Houston.

Advocates for data-driven methods complain about the barrage of buzzwords used by consultants promising huge improvements but have little experience in the oil business. And they need to sort out which technologies are dependable enough for running a business.

Deep learning is on the list of promising but unproven technologies. It is a powerful emerging technology capable of solving some really difficult problems, such as facial recognition.

“People need to be very wary of these techniques,” said Richard Baraniuk, professor at Rice University, who is working on building elements into deep learning to understand why it works to solve some problems but not others.

“Five years from now we will be able to categorize which method works better. But we are not there yet,” he said.