By Jim Redden; World Oil Magazine; June 2019 (Vol 240 No. 6)
With apologies to writer and philosopher George Santayana, well programmers who have no insight into the drilling problems of the past are condemned to repeat them. When designing an offshore exploration well, it would be nice to know, for instance, how drilling teams managed to remediate chronic lost circulation or stuck pipe incidents in a certain interval of offset wells in the area.
Getting to that data typically means sifting through thousands of data sets—even if they can be found—resulting in countless hours of digging and reading, and the likely consumption of prodigious amounts of antacid. This tedious process can be particularly befuddling to millennial-aged engineers more accustomed to fingertip searches than sorting through filing cabinets.
A company in California’s Silicon Valley claims to have made the data gathering and alignment process much more eyesight-friendly, with what it says is the industry’s first knowledge graph for wells. Driven by artificial intelligence (AI), the new WellLine application enables “engineering and operations staff to identify risks and best practices from historical operations.” Describing itself as a pioneer in digital knowledge technology, Maana, Inc., says WellLine runs on top of the Microsoft Azure cloud computing service to mine millions of drilling reports, operations reports, and other well-related data across thousands of offshore and onshore wells, “securely and at scale.”
“You have wells, and something happened. People have to read through all this data, and that’s just crazy. The goal here was to take data, turn everything into an event and create a timeline. We want to recreate a history of the well from the perspective of time. So, you bring up a little window, type in the name of the well and you get everything that ever happened on that well,” said Jeff Dalgliesh, director of oil field digital transformation, on the sidelines of last month’s Offshore Technology Conference (OTC) in Houston, where he presented the paper, “AI-Driven Well Timelines for Well Optimization.”
The former Chevron hand says making the historical data search manageable is all about mitigating risks. “When you’re going to drill a well, you have to figure out how to mitigate risks. We mine all this historical data and put it in a place engineers can use, to see what risks they should be mitigating against. What kind of problems did people have in an area?” he said.
Maana, which counts Chevron Technology Ventures, Saudi Aramco Energy Ventures and Shell Technology Ventures among its eclectic group of investors, completed the first phase of the WellLine development and beta testing in January, and is now in the roll-out stage. While the platform comprises look-back data that take in every operation, from exploration to abandonment, development actually began at the end of the well life cycle.
“When I first started, I wanted to make it easier to do an abandonment,” Dalgliesh said. “You go into abandonments, and you have data everywhere; you have texts in different formats and in different places. So, how could you program an algorithm to go through and mine all that historical data?”
Much of the early development and training of the WellLine application relied on some 4,000 dataset files that Equinor ASA provided upon the 2016 abandonment of the 23-well Volve field in the Norwegian North Sea. The software, likewise, includes U.S. Bureau of Ocean Energy Management (BOEM) data, comprising some 3 million drilling comments on more than 46,000 wells throughout the Gulf of Mexico. All data are organized and indexed around common oilfield concepts, namely equipment, vendors, well activities, drilling problems, HSE events, incidents, and lessons learned. The data can be filtered to the extent that comments a specific drilling manager may have made, on a certain section of a certain well, can be accessed by simply inputting the person’s name.
“We index everything against common oilfield terms, like stuck pipe and lost circulation, and we include every piece of equipment we can get our hands on,” he said. “We now have around 1,500 pieces of equipment, so we can pull, for example, anytime anybody has mentioned BOP and testing.”
Dalgliesh says relying on new generation sensors and the like for data collection definitely had its place, but capturing and making useful sense of historical data is another matter altogether. “We’re talking about going after all this historical data that goes back to the early eighties and even current stuff. A lot of NOCs (national oil companies) inherited data from different companies, and it’s all in different formats and they have files of it,” he said.
“We wanted to build something that you could just search. You can type in any search term, even something as specific as H2S (hydrogen sulfide), and ask to show every drilling comment referencing H2S in a certain area and give me a map of it. Or, put in lost circulation and give me a map of it, and at what depth did it happen.”
Though WellLine was built strictly to capture and organize historical data, Dalgliesh said integration work is underway with companies specializing in predictive algorithms, to allow clients to include a prediction timeline. “There’s a lot that can be done with all the new AI stuff that’s out there. Operators can see when production in a well might start to decline or when a piece of equipment might fail.”
The bottom line, Dalgliesh says, is that companies can now access potentially valuable data that may otherwise remain forever locked in a dusty file or in the mind of a retired engineer. “There’s a huge amount of data out there that’s basically ignored.”