Machine learning is transforming the way in which oil and gas is discovered and developed. And the help of big technology companies has catapulted the oil and gas industry toward a new course of action.
Just a week ahead of the 2018 Oil and Gas Machine Learning Symposium, Rigzone caught up with speakers John Adamick, senior vice president data and analytics for TGS; Dr. Christian Noll, geoscience manager, advanced analytics and emerging technology for Anadarko Petroleum Corporation; and Dr. Tom Smith, president and CEO of Geophysical Insights, to discuss how machine learning and other technologies are changing the oil and gas industry.
Rigzone: How is Anadarko currently utilizing machine learning within its organization?
Noll: Machine learning is a key component to our digital transformation at Anadarko. Within AAET Geoscience, we are using machine learning (ML) techniques to enable our exploration and development success in the deepwater Gulf of Mexico through the use of deep learning networks to detect features such as faults, salt and stratigraphic geometries, giving our seismic interpreters a sophisticated ML-based workflow to both accelerate their interpretation and more importantly provide a robust feature detection tool with a more exhaustive approach than manual efforts can provide in isolation. To enable our Lower 48 teams to improve subsurface characterization efforts, we are injecting ML-based techniques into workflows to clean and enrich the density of our onshore well-log datasets to make better use of the data we have. In addition, we are using ML-based tools to help our geoscientists build stratigraphic frameworks, which facilitate our petrophysical and map-based evaluations, enabling us to screen potential opportunities rapidly for greater optionality.
Rigzone: Technological efficiencies adopted during the downturn by numerous oil and gas companies have cut down on the need for workers to handle some tasks. What do companies expect from their employees now? What skillsets are desired?
Noll: First, it’s important to recognize that machine learning doesn’t necessarily replace people, it enables our talented people to work better, safer and smarter. We see the adoption of ML-based interpretation solutions as the next evolutionary step, similar in nature to the transition from 2D line-based hardcopy seismic interpretation to early workstations in the past, or the switch from early 2D workstation interpretation to more advanced 3D interpretation toolsets. As before, the geoscientist of tomorrow will still need strong core geoscience skills, but to embrace these next-generation ML-based toolsets and thrive in the revised landscape, a hybrid skillset of core geoscience along with strong mathematics skills, coding capability, creative problem-solving and an ability to communicate across domains will be increasingly useful skills in the geoscience profile.
Rigzone: In a time where companies are keener on using machine learning in the E&P space and hiring more workers who are adept at it, how can companies ensure they’re providing quality data for machine learning engineers to work on?
Adamick: Machine learning algorithms require both quality and standardized data to yield the best results. Data companies like TGS work very hard to deliver on both. For seismic, this process begins with data as it is being captured in the field, but also includes lots of back office activity to ensure that both the data and metadata are standardized and in sync. In the case of historical well data, you often run across the case where some data was not ever collected in the field. TGS is developing machine learning techniques to infill this missing data and provide our clients standardized well data sets ready for use with their own machine learning programs.
Smith: Collecting and maintaining quality data has been a long-standing challenge that our industry has faced. The engineers, geoscientists and data scientists using various analysis tools, including machine learning, are well aware of the impact that data quality has on analytical results. We view the problem of data quality somewhat differently in that we see machine learning technology helping to identify noise in the data and revealing limitations of a given set of data.
All data scientists realize that subject matter experts are required for proper analysis, so it is not just a matter of throwing data at the machine. The machine learning data scientists need geoscientists and engineers to understand and vet the data and likewise, the subject matter experts need data scientists for the development and understanding of these new complex algorithms.
Rigzone: What can the oil and gas industry learn from big technology companies?
Adamick: The oil and gas industry can be rather provincial and sometimes slow to adopt new ideas. The big technology companies are the opposite. They move fast, try out many new ideas and leverage their expertise over many different industries. The oil and gas industry could benefit from adopting some of these traits, especially by learning from processes that have worked in other industries.
Smith: Big Tech has invested billions in new technologies such as cloud computing, high-performance computing (HPC) and graphical process unit (GPU) computing. While these companies are generalist, not specialist, when it comes to the oil and gas industry, the technologies they bring to the table can be leveraged by geoscientists, engineers and data scientists who are oil and gas domain experts. It’s also important to keep in mind oil and gas data is different than consumer data. It’s incredibly expensive to obtain, the decisions taken based on the data have significant upfront capital requirements and the data is driven by the fundamentals of physical principals.
Noll: The increased footprint of onshore unconventional activity in our industry has brought with it a reliance on rapidly increasing datasets, which puts pressure on our existing technology from decades past and our ability to efficiently handle and quickly process these very large datasets. Big technology companies rapidly evolve with the demands of a changing landscape and are increasingly reliant on the agility of elastic solutions to scale up or down to meet this demand. In addition, these companies empower their highly skilled workforce to embrace creative solutions to challenge the existing paradigms and embrace solutions from outside of the mainstream of their given domain. These are all enabling foundations that will help our own industry navigate the opportunities and challenges that will present themselves in the future.