ISI Logo

Domain Expertise and Data Analytics in the E&P Industry

Shahab D. Mohaghegh - June 2017

Domain expertise is an absolute necessity for successful, meaningful, and practical application and implementation of data driven technologies (Data Science) in the oil and gas industry. It may be argued that domain expertise is the source of what some choose to call preconceived notions and biases held by engineers and geoscientists that data driven technologies are set to overcome and substitute with facts, field measurements, and realities. To support this claim, some statisticians, AI advocates, and Silicon Valley entrepreneurs (with minimal to no domain expertise in the oil and gas industry) cite the lack of success, or the minimal value added to shale development that is made by traditional petroleum engineering approaches (relative to the investment made in these technologies) such as Numerical Reservoir Simulation (NRS), Rate Transient Analysis (RTA), and Decline Curve Analysis (DCA). Regardless of their popularity, and applicability to conventional resources, these techniques have provided minimal insight into the complexities associated with the completion and the production from shale wells.

While there is a lot of truth in the above assessment concerning the application of traditional petroleum engineering techniques to production from source rocks, there is a problem with this line of reasoning. There is a fundamental difference between the required “Domain Expertise” and preconceived notions and biases. They should not be confused. While “Domain Expertise” are among the absolute essentials in the success of incorporating data driven analytics in the oil and gas industry, preconceived notions and biases are the greatest obstacles and huddles that our industry need to overcome. However, they are not the same thing.

“Domain Expertise” is what some statisticians, AI advocates, and Silicon Valley entrepreneurs lack and is the main reason behind (at best) the mediocre results that they have generated (and some are still generating) in their efforts to bring AI and Data Science to the E&P. Domain expertise provide the type of insight that can result in innovation, once new technologies are presented. Domain expertise is the depth of knowledge that recognizes substance when it is exposed to innovation and can identify opportunities in the given field. Domain Experts, are familiar with the short comings of the existing technologies and by definition are open minded individuals that do not shut the door on anything new that they are not accustomed to. In the case of “Engineering Application of Data Science,” petroleum engineering domain experts that are well versed in the art and science of data driven solutions are able to build the required solution architecture that guides the data toward desirable, meaningful results. This is one of the main differences between using data driven solutions for engineering problems versus using it in the analysis of data collected from social media or finding solutions and analyzing the type of problems with no physics behind it. Not only the predictive analytics, but also the exploratory analytics applied to data generated from physics based phenomena are fundamentally different from those applied to data generated from non-physics-based problems such as data relating to human behavior or social issues.

The solution architecture that results from the domain expertise, incorporates the understandings related to potential values of different steps in the operations and their general (high level) potential contribution that must be included in the process, in order to make the data driven solutions meaningful. These domain experts are the ones that are able to connect the dots by using the capabilities of the data driven technologies to overcome the short comings in our traditional solutions and to develop completely new ways of addressing problems that have been eluding our engineers and scientists and have left us with no deterministic solutions for certain problems. The complexities associated with such problems (production from shale being a good example) stems from our lack of required insight into the details of the physics necessary to completely control and optimize these physical processes.

Preconceived notions and biases, on the other hand, result from a lack of deep understandings of the essence of the physics that is being observed. They are the lazy and the superficial extensions of anecdotal evidences and observations to general and fundamental physics of complex phenomena. Those that fear change and feel threatened by any alteration in traditions, develop preconceive notions and biases that they stick to, no matter what the reality puts in front of them. To some of these individuals, it is only the geology, the petrophysics, and the geophysics of the shale that drive production, therefore, in their view you should be able to copy-paste the completion design from well to well and it would make no difference in production and recovery. On the other hand, same type of individuals with identical characters but completion expertise (or lack thereof), believe that the way the well is completed and stimulated is the only reason behind its production behavior, since what we do to the source rock (design parameters) are the controlling factors and are far more important than the source rock’s natural characteristics. Traditions form preconceived notions and biases.

The fact is that historically, and prior to the advent of data driven analytics and modeling as a mainstream predictive technology, domain expertise of the engineers and geo-scientists needed to be formulated into a set of deterministic mathematical equations. The equations were then solved analytically, or numerically, to form the set of solutions that have been used for decades in our industry. Pressure Transient Analysis (well testing) and numerical reservoir simulation are examples of this traditional historic approaches. One of the most exciting contributions of the data driven analytics to the field of engineering is that formulating the complex physics into a set of complex, non-linear, partial differential equations and then using mathematics to solve it, is no longer the only viable solution. We now have another potential approach to address such engineering problems.

The traditional approaches worked well during the past century when limited amounts of data were collected. The majority of the data that we used to collect were limited to well logs, and core analysis from the subsurface and pressure and production volumes from the surface. Even when we had large amounts of data from sources such as seismic surveys, we had not been able to develop deterministic technologies that were able to fully integrate them into our reservoir and production engineering workflows.

The world has changed. Using the newest sensor technologies, we can now collect massive amounts of data (Big Data) from places down-hole that we never had continuous access to, in the past. All the previous technologies and our dependence on traditional physics-bases approaches stemmed from the fact that we had to, to a large extent, rely on guessing what is actually happening in the wellbore and in the reservoir as we drilled, completed, and produced hydrocarbon. Of course, from time to time, some operators (with extensive resources operating prolific reservoirs), were able to perform certain types of tests to generate anecdotal evidences to help them in their endeavors, however, these anecdotal evidences (tracer tests are good examples) are not enough, or always available.

We are now living in a world where we can collect massive amounts of different types of data with high resolutions in time and space, using down-hole measurement techniques. These data has already overwhelmed our capabilities to analyze them, let alone to model them as a predictive tool to help us control and optimize our operations. We now have access to these facts, field measurements, and data. Furthermore, we do have the domain expertise and the AI and Data Science expertise to utilize and maximize their potential contribution to our day to day operations. Shale Analytics and Real-Time Shale Analytics makes this possible and will turn this potential into reality. It is only a matter of time.