With the recent advances in technology and adoption of machine learning in geosciences, is it time for AI to take a hand in estimating hydrocarbon prospect sizes?
Comparing company pre-drilling resource estimates for oil-in-place with post-drilling discovery size reveals a significant misfit. The green lines show the P10–P90 range, the line of squares are the expected discovery size pre-drilling, while triangles represent the estimated size post-drilling. Source: NPD.
Explorationists are optimists. It’s a proven fact. It must be part of their genes.
The truth is that they just cannot avoid overestimating the size of their prospects. It has been said before, and it was reiterated at the NCS Exploration Strategy Conference in Stavanger in November.
The Norwegian Petroleum Directorate has statistics from several decades of exploration on the Norwegian continental shelf (NCS). Hans Martin Veding, statistician at the organisation, referred to 33 wells drilled from 2015 to 2017 that had drilling targets with comparable pre- and post-drill estimates. Only two (!) of them found more than was originally projected; the remaining 31 were decidedly overestimated. Veding showed the same to be true for prospects drilled in acreage allocated from the 8th to the 22nd round on the NCS. Prospects were massively bigger than the discoveries.
Paul Herrington, Exploration Portfolio Manager with the Oil and Gas Authority in the UK, reported the same observations from the UK sector. Herrington based his statistics on 750 fields and discoveries across the UKCS since the first well was drilled in 1965, and there is absolutely no doubt that the oil companies are consistently far too optimistic.
One might speculate that without such overrating, many wildcats would not have been drilled, resulting in fewer discoveries.
However, the oil companies are not alone in making false predictions. Graeme Bagley of Westwood Global Energy Group referred to studies by USGS that show pre-drill resource estimates for sedimentary basins around the world. Out of some 20 selected basins, USGS has been wrong in all of them, except for the offshore Guyana Basin where ExxonMobil is in the process of proving up some 10 (to 20?) billion barrels of oil, possibly more. In all the other basins investigated by USGS, the actual volumes are far less than the USGS P10 estimate. The USGS screening cannot therefore be judged as meaningful.
Given that USGS does not have access to lots of data, this is perhaps easy to understand. For the oil companies, however, there must be another reason. Unsupported optimism is here suggested as one possibility that very few would argue against. It would be interesting to know if using artificial intelligence – on huge amounts of data – could help explorationists come up with more realistic estimates. Should the oil companies possibly rely less on the human brain and leave it to computers to take history into consideration?
Further Reading on Artificial Intelligence in Oil and Gas
Some recommended GEO ExPro reading relating to, or similar in content to, the use of artificial intelligence in the exploration and production of oil and gas resources.
Artificial Intelligence in Oil & Gas Production
Bjørn-Erik Dale & Vidar Uglane; Solution Seeker AS
Solution Seeker, a Norwegian tech start-up and spin-off from the Norwegian University of Science and Technology, is developing the world’s first artificial intelligence for real-time production optimisation.
This article appeared in Vol. 15, No. 3 – 2018
Part I: An Introduction to Deep Learning
Hongbo Zhou, Statoil; Lasse Amundsen and Martin Landrø
Once, artificial intelligence (AI) was science fiction. Today, it is part of our everyday lives. In the future, will computers begin to think for themselves?
This article appeared in Vol. 14, No. 5 – 2017
