The Digital Future of E&P Can Lead to Data-Driven Anxiety
The constant talk about the data-driven future of the oil and gas business poses a threatening question for some petroleum engineers: What do I need to know to ensure I have a job next year?
Many universities are adding digital data and analytics programs to prepare petroleum engineering students, many of whom are also taking the initiative on their own to master the tools used for this new way of working in the industry.
Jim Crompton, an adjunct faculty member at the Colorado School of Mines who created and taught some of the first such classes, said students who grew up in the Internet era pick it up quickly. He said he worries, though, about working engineers.
“The greater challenge is for those with 10–20 years of experience,” he said, specifically engineers who do not know programming and do not have the vocabulary of digital analysis.
Engineers who survived the mass layoffs of the last downturn are likely to be suffering from digital anxiety, he said. Some have seen warning signs. “They didn’t get the job they wanted or a promotion or something like that,” Crompton said. A major hurdle for many of those engineers is that their demanding jobs leave little time for training.
Managers are more likely than others to be unaware of the need to know digital-data concepts, Crompton said. Decision makers also need to understand digital-analysis methods well enough to get a feel for whether the analysis is legitimate. Failure to detect flawed analysis when approving projects can “waste a bunch of money,” Crompton said.
As digital-data programs proliferate, SPE is developing an online training program and is working on curriculum guidelines for data science and digital engineering in petroleum engineering schools. Birol Dindoruk, SPE’s technical director for management and information, who has made data-related issues a priority, said he expects to submit the guidelines to the SPE Board of Directors this year.
Need To Know
Making the case for knowing about digital data is relatively easy, but figuring out what a petroleum engineer needs to know can be complicated. The amalgam of petroleum engineering, data science, and information technology has so many elements that finding a short label for it is difficult. Requirements also vary on the basis of job descriptions.
One thing is clear: Collaboration skills are required. Companies with data scientists, often from outside the oil business, tend to pair them with an engineer. The range of knowledge those two possess exceeds what each is expected to have, Crompton said.
Engineers managing assets will need to have a working knowledge of the tools and vocabulary used to collaborate with data engineers, and a strong base of traditional engineering concepts is needed because advanced data analysis can generate multiple answers, some of which are unhelpful. “When it is wrong, it can miss by a mile,” Crompton said. Artificial intelligence is “a bit of a leap forward into the unknown,” he added. A person with a firm grasp of the physics of oil and gas exploration and production is required to help identify the good and bad ideas.
“People hiring (petroleum engineering) students do not want to hire data scientists. They want to hire petroleum engineers who are conversant and capable in these areas,” said Jon Olson, chairperson of the Hildebrand Department of Petroleum and Geosystems Engineering at the University of Texas at Austin (UT).
The amount of training options available from colleges, SPE, and some oil companies is growing. Still, no hard definition exists for how much an engineer needs to know.
On a recent visit to the Marietta College Petroleum Engineering Department, Jeff Moss, SPE’s technical director for drilling, asked faculty members, “With advanced analytics, is that the domain of petroleum engineering itself or is it a tool to be used in that domain?”
In other words, how much should an engineer be expected to know and do? Writing the code for an advanced analytics program may be asking too much. But, when a significant decision relies on analytics, petroleum engineers likely will need to be able to conduct a quality check to evaluate the process used.
When asked how he would answer the question, Moss said, “I was stirring the pot. I don’t know the answer.”
Another consideration, said Lloyd Heinze, a petroleum engineering professor at Texas Tech University, is “we are using a computer program and are not sure what it is doing.”
Finding out how the computer is arriving at conclusions requires knowing enough about the inputs, the algorithm, and the processing engine and being able to check the results with petroleum engineering fundamentals. Students must be able to “derive an equation and look at some real data and see if the equation predicts the data,” Olson said. If it does not, he said, “Where do you go from there.”
Educators can have trouble keeping up when adding digital-analysis training at a time when rapid advances in producing oil from terrible quality reservoir rock—unconventional reservoirs—expands what engineers need to know.
“While a course in data analytics and how to use current software or control-system programming would be interesting and potentially of value in the near term, the physics of the drilling process and its limiters is a topic that will have the longest lasting impact on a student,” said Sam Noynaert, an assistant professor at Texas A&M University.
Colleges need to adapt with the times, but they are never going to be able to keep up with the software business. “The technology, programming languages, and other aspects evolve so rapidly that the curriculum is outdated the first time it is taught,” Noynaert said, adding, “What is successful is teaching the underlying fundamentals, as these will not change.”
University programs are designed to minimize the time spent learning how to use data-analytics tools so students can quickly begin using them to solve problems. The objectives for the Colorado School of Mines course include applying the principles of data analytics, with projects that include interpreting the performance of thousands of wells and building a business plan.
The classes require students to use Spotfire, a powerful data-analysis and visualization tool widely used to analyze large data sets. “Doing their homework with Spotfire allows them to build proficiency,” Crompton said, adding, “They can put that on their resumes,” which will mean they possess a skill many field engineers do not.
Other universities follow a learn-by-doing approach, although the tools used vary. At Texas A&M, Matlab software is used in classes for analysis and visualization. “I have seen many students who realize that Excel is a nice but limiting tool,” Noynaert said. “Since they are engineers, many are starting to use Matlab for some types of work since we use it in some reservoir engineering courses already.”
Adopting New Tools
At UT, Olson said the emphasis is on programming in Python and R, which are widely used for analytics. Multiple sources mentioned that advanced users, such as graduate students, are likely to gravitate toward those languages, sometimes signing up for private boot camps to learn them.
“In the last 1 to 2 years, I have seen an exponential increase in students who are learning R and Python on their own. This indicates that students realize that, in order to manipulate, analyze, and present large amounts of data, new tools are required and it is worth their time to learn about these tools,” Noynaert said.
At UT, the petroleum engineering department will soon offer an elective course on subsurface machine learning, Olson said.
For one class offered as part of the university’s Rig Automation and Performance Improvement in Drilling (RAPID) consortium, research assistants are hired to analyze big data sets from participating service companies and operators. At the end of the semester, the students advise the companies on the basis of what they observed, said Pradeep Ashok, a research scientist in the university’s Cockrell School of Engineering.
“For the students, working on these real-world data sets and problems is many times more beneficial than just studying statistics and machine-learning theory,” he said. “Here, they apply skills learned on real data.”
Working engineers face the challenge of finding the time to solve engineering problems while keeping up with growing responsibilities and the data that come with it.
“The bar is rising for what the industry expects from petroleum engineers,” Crompton said, adding that machine assistance is required to keep up when business plans call for engineers to do twice as much work without adding staff.
Training at work is increasing, too. “Everybody is doing digital, and continuing education of mid-career petroleum engineers (and other engineering disciplines) is a focus,” Crompton said. He said the list of companies with in-house digital-data training includes BP, Chevron, ConocoPhillips, Anadarko, and Pioneer.
To help fill this need for those with such options, SPE has launched its first online course on data science and digital engineering. Crompton is the instructor for the three-part course, which offers an overview of the digital oil field, a review of data management and analytics, and a look at field applications.
Texas A&M recently launched a distance-learning master’s degree in statistics “that is attracting a lot of interest from engineers in all industries,” Noynaert said.
UT is planning to start a digital-data-analytics boot camp for working professionals this summer, Olson said.
Petroleum engineering schools are responding to advances in technology, and industry pressure, by revamping their courses to keep up with digital and unconventional changes.
The SPE Petroleum Engineering Education Colloquium in August saw calls for more training on petroleum engineering fundamentals—particularly updates on recent advances— plus more training on data and analytics.
Good arguments can be made for why both of those are important, and also why they will be hard to deliver. Enrollment in petroleum engineering programs is down, reducing tuition revenue. Also, petroleum engineering programs already are at the limit for the number of required hours and faculties may lack staff able to teach courses such as data and analytics, said Ramona Graves, SPE’s technical director for academia.
Given the limits, academics are trying to figure out which of the requests is the highest priority; but, Graves said, “It is hard to get a consensus.”
Olson said the digital-data experimentation at multiple schools is likely to progress toward similar endpoints. Olson pointed out that many of the same big companies are on the schools’ industry advisory boards and that the department heads meet regularly to share ideas.
The petroleum engineering program at Louisiana State University (LSU) replaced a single course in probability and statistics with a two-course sequence that included data analysis and visualization, programming, and numerical methods, said Karsten Thompson, department chairperson for the program. The university is also adding an analytics elective for seniors.
LSU was able to revamp what had been just a statistics course because it had a teacher on staff with the necessary expertise. Students there can draw on other resources—for example, those interested in automated drilling can take classes on digitally controlled operations in the mechanical engineering department.
While the expanded data courses currently are electives, they will be required in the future. This mandate is part of an overall review of LSU’s petroleum engineering curriculum that added more content on topics related to unconventional reservoirs—such as rock mechanics—and more on designing completions in those difficult formations. LSU also added an elective on directional drilling.
Adding new offerings raises the question of what can be cut. Thompson said that LSU has made time for new material by updating some sections and eliminating overlapping material.
At the Colorado School of Mines, creating an oilfield-data elective required recruiting Crompton. He fit the bill because of his career experience with digital oil fields at Chevron. He also helped create and continues to contribute to a class on the subject at the University of Southern California (USC). He said his Colorado roots made moving to the Colorado School of Mines an attractive option after he retired from Chevron.
The undergraduate class there is different than the one at USC for graduate students, whose research work requires the ability to perform advanced analytics.
Digital data emphasizes probability-based approaches, while petroleum engineering math requirements strongly tilt toward calculus. Heinze said that Texas Tech students take four calculus classes and then choose between statistics or linear algebra. Given the importance of statistics and probability, he said, “I am wondering if one statistics class is enough.”
The Norwegian University of Science and Technology (NTNU) has made a concerted effort to increase its focus on digitalization and automation in petroleum engineering, said Alexey Pavlov, a professor of petroleum cybernetics at NTNU.
Pavlov’s job as cybernetics professor is part of a collaboration between the Department of Geoscience and Petroleum and the Department of Engineering Cybernetics at NTNU, which is funding research applying the study of communications and automatic control systems to data-driven, digital control systems used in oil and gas. “One needs to know multiple disciplines and understand interplay and interfaces between them,” Pavlov said. He pointed out that “specialists with both a traditional petroleum engineering background and digital and automation expertise are, and will be, in high demand in the near future.”
Don't miss out on the latest technology delivered to your email monthly. Sign up for the Data Science and Digital Engineering newsletter. If you are not logged in, you will receive a confirmation email that you will need to click on to confirm you want to receive the newsletter.
09 May 2019