January 2024

What's new in exploration

2024: The two lies of exploration
William (Bill) Head / Contributing Editor

Lie 1. “AI, AI, and it’s off to work we go…” When you push for AI, are you telling your company that your job is no longer necessary? Let me help. AI finds patterns, y to x, find z along the x-y line or project past either x or y, no data needed, and that’s it. Once a set of data is entered, a program can only choose avenues that were previously identified. AI can still not convince us that the blue car you see every Friday at noon is blue, because it is Friday or noon. We are not quite at the place where AI takes over oil and gas, but I wonder if AI could run regulators, assuming that would be a benefit [provided I could tell AI what is preferred]. Woe to us, if the folks in control gain an understanding of how to program in C++. 

What happens when AI predicts an output that another AI “believes” as data input for further calculation, say of permeability, porosity, pressure, or tomorrow?  At this stage in an algorithm, AI is much more involved than a mere iteration of unknowns reducing to a set finite difference. Recall that Deepwater Horizon’s crew used early AI models to monitor their blowout preventer. Did anyone tell the pre-AI computer that not all four valves were in service, then adjust? Remember, garbage in, garbage out. 

Today’s well-logging AI computer analysis “elements” ignore actual tool calibration when compared to prospective reservoir rock. Calibration is assumed dependent on bench tests of electronics and metal that actually very few on-site logging engineers have ever used with their (your) equipment at an actual test pit. Was there a calibration inference test using on-site radioactive materials?   

Fig. 1. Cross-plots from a formation evaluation lecture in a petroleum engineering class at Texas A&M University. Source: https://blasingame.engr.tamu.edu/z_zCourse_Archive/P663_10B/P663_Schechter_Notes/PETE_663_CROSSPLOT_B_DSS.pdf

During the old analog days, a well-log operator would have film strips of “calibration records” to add to paper log prints, as if that were true. Once, I hired 3 well-known service companies to log a 4,000-ft well. Each had a different calibration protocol; none were on the spot, with actual calibration devices common to the practice. No calibration agreed with my supplied science-grade molecules. Simple mud tools can drastically impact the use of Rwa in correcting mud invasion for e-logs. E-log water-cut has been everything in oil prospectivity, followed only by log “cross-over.” Oh, how quickly we forget the basics. But beware of good enough in, but dry hole out. Recall the cross-plots from Blasingame’s class at Texas A&M in Fig. 1. Perhaps you should, or at least understand, that AI has yet to replicate a common decision cross-plot, where one x,y data point could be caused by a series of phenomena.  

A rather decent exposure to the benefits of data cross-plots can be found in this Journal of Petroleum Exploration and Production Technology article: “The use of cross-plots in lithology delineation and petrophysical evaluation of some wells in the western Coastal Swamp, Niger Delta,” Okwudiri A. Anyiam1 • A. W. Mode 1 • E. S. Okara 1 J Petrol Explor Prod Technol (2018) 8:61–71.

I see a similar misinformed attitude today among seismic interpreters, who believe AI-interpreted 3D does not have reflection interbed multiples. Hmm. What are geos telling the AI programmers? Fantastic reflectors exist offshore GOM-Florida, Western Australia, and Qatar that completely mask the subsurface. Some occur significantly above massive oil and gas reservoirs. You can still see those “events” on 3D seismic data, often marketed on trade show banners.

Lie 2. All the good stuff has been found. Maybe from hindsight or the distraction of M. King Hubbard’s 1956 peak theory, it may appear that all the big reserves have been found. Technology changed that idea with the analog to digital 2D seismic revolution (1964); then again, technology exploded oil and gas exploration with 3D seismic (1973, Seisloop at G.S.I., and various geometries popularizing 3D in the early 1980s). While a swimming pool is finite, the earth’s pool size is still unknown. George Mitchell, et al., added some common sense when advancing the fracing method (1997) as a big deal in the Barnett shale. However, the shale-fracking apocalypse (2007) ignited a new hate for fossil fuels. 

Of course, not to miss an opportunity to profit from other’s political woes, we seem to be losing some of that common sense when world governments tell us from their luxury seats on private jets to abandon fossil fuels. Good news! Not everyone is convinced. Congratulations to Exxon and Shell for going public on the future of fossil energy while continuing efforts to buy other energy companies and to find new reserves. Share price does matter! Note that the Saudis are still exploring Ghawar, the largest known oil field in the world (to date). First discovered in 1948 using an anticlinal surface feature, in spite of sand. What we don’t know, we don’t know. 

About the Authors
William (Bill) Head
Contributing Editor
William (Bill) Head is a technologist with over 40 years of experience in U.S. and international exploration.
Connect with World Oil
Connect with World Oil, the upstream industry's most trusted source of forecast data, industry trends, and insights into operational and technological advances.