New Tech Breathes New Life into Well Logging

Basic. Humble. Old school. Whatever you call them, petrophysical well logs are a workhorse for the oil and gas industry.

Today, geoscientists are using machine learning and other artificial intelligence techniques to glean even more information from logs. And a new generation of downhole instruments now provide loggers with more, and more relevant, data.

“Well logs are our best tool for getting a continuous look at the rock from the bottom of the well to the top. There’s just no cost-effective way to do continuous coring for the whole well,” said Mike Maler, chief geoscientist for Enervus in Austin.

Enverus, a data analytics company, changed its name from Drillinginfo in August in a rebranding move. It had recently discovered a chance to quadruple its interpreted well log catalog – and jumped on it.

Shortly after the name change, the company announced it had acquired about 200,000 geological well datasets from Reservoir Visualization Inc.

The RVI log acquisition “was key to Enverus,” Maler said. “It gave us a lot more spatial coverage of both well logs and correlated formation tops.”

“We’ve had geologists working on the Permian Basin for several years, and there’s still work to be done. We continue to refine our thoughts about it. Adding this data, we can continue to study it for several more years,” he added.

In well-log analysis, the geoscientist’s toolkit has expanded to include machine learning, deep learning and other AI-aided approaches to assess data. Machine learning draws on advanced computing power to identify patterns in data and to make basic decisions based on that data.

Deep learning is usually considered a type of machine learning that utilizes the concept of artificial neural networks, inspired by and based on the way the brain works.

Please log in to read the full article

Basic. Humble. Old school. Whatever you call them, petrophysical well logs are a workhorse for the oil and gas industry.

Today, geoscientists are using machine learning and other artificial intelligence techniques to glean even more information from logs. And a new generation of downhole instruments now provide loggers with more, and more relevant, data.

“Well logs are our best tool for getting a continuous look at the rock from the bottom of the well to the top. There’s just no cost-effective way to do continuous coring for the whole well,” said Mike Maler, chief geoscientist for Enervus in Austin.

Enverus, a data analytics company, changed its name from Drillinginfo in August in a rebranding move. It had recently discovered a chance to quadruple its interpreted well log catalog – and jumped on it.

Shortly after the name change, the company announced it had acquired about 200,000 geological well datasets from Reservoir Visualization Inc.

The RVI log acquisition “was key to Enverus,” Maler said. “It gave us a lot more spatial coverage of both well logs and correlated formation tops.”

“We’ve had geologists working on the Permian Basin for several years, and there’s still work to be done. We continue to refine our thoughts about it. Adding this data, we can continue to study it for several more years,” he added.

In well-log analysis, the geoscientist’s toolkit has expanded to include machine learning, deep learning and other AI-aided approaches to assess data. Machine learning draws on advanced computing power to identify patterns in data and to make basic decisions based on that data.

Deep learning is usually considered a type of machine learning that utilizes the concept of artificial neural networks, inspired by and based on the way the brain works.

“The toolkits are getting better, but we are still addressing the same problem, which is to be predictive about what your next well is going to see or how it will perform,” Maler noted.

“We used to use Excel spreadsheets. Now we use machine learning. We’ve been using multivariate statistics, as well,” he said.

AI and advanced computing enable geoscientists to model the subsurface by applying huge amounts of data from collections of well logs, which also serve as a reference check.

“We build a model and say, ‘That’s not fitting the data,’ so we create another model almost in real time,” Maler said.

Supervised Learning

That reality-check aspect of log information also helps fill out and verify the subsurface picture when techniques like machine learning are applied to limited datasets, according to Jimmy Fortuna, senior vice president of products at Enverus.

“With the advent of powerful and inexpensive machine-learning technology, some (companies) have statistically modeled the subsurface using sparse datasets,” Fortuna noted.

“Models produced using machine learning or statistical methods are only as good as the ground truth datasets used to train them,” he said.

That’s not just another way of saying “the more data, the better.” Maler emphasized the importance of adequate data examples used to “train” computers in machine learning, a process known as supervised learning.

“Whether you want to talk about machine learning, deep learning or multivariate analysis, from a philosophical point of view the key is: How good is your calibration or training data?” Maler said.

“If you haven’t invested the time and effort to build a high-quality dataset you can train off of, you’re sort of wasting your time. There’s no shortcut here,” he observed.

Advances in Downhole Logging Tools

Service companies have responded to the challenge of unconventional reservoir analysis and characterization with a new generation of downhole logging tools. And “X” turns out to be an important letter.

Last year, Halliburton introduced its Xaminer Magnetic Resonance Service, also called XMR, billing it as “new-generation combinable wireline magnetic resonance.”

The XMR sensor delivers detailed formation data including 2-D and 3-D fluid characterization, carbonate pore-size classification, unconventional analysis and permeability readings, Halliburton announced.

It said XMR can acquire about eight times more data with less than half the power of traditional sensors.

Baker Hughes has introduced its Array Dielectric eXplorer formation-evaluation service to provide key petrophysical data to improve the quality of reserve estimates.

According to Baker Hughes, the eXplorer service relies on dielectric permittivity data to identify hydrocarbon saturation in reservoirs with any water resistivity, using a dual-resolution sensor array to simultaneously acquire a large volume of data at a common measure point.

Petrophysics Comeback

The advent of unconventional resource plays brought petrophysics back to the forefront of the oil industry’s geological studies, according to Maler.

“It made petrophysics relevant again almost overnight, or so it seemed. Hiring and retaining petrophysicists became a huge challenge,” he recalled.

And in petrophysics, also, machine learning and other AI techniques are helping geoscientists understand and characterize the subsurface.

“In the past few years I’ve seen a growing acceptance of, and more and more people using, machine learning to accelerate their petrophysical evaluations,” Maler said.

“Now you see people using neural networks to try to accelerate that, using (well log data and interpretations from) 12 to 15 wells and applying that to 100 wells. It will be interesting to see where that goes,” he added.

Petrophysical evaluation looks repetitive “in a superficial way, but it really isn’t,” Maler noted. “Like geology, it’s more of a pattern-recognition, interpretive science based on fundamental principles.”

Expanded Dimensionality

With more and better data and improved tools and techniques, well logging remains an essential part of the big picture in exploration and production. According to Maler, that big picture now has considerably more dimensionality than it once did.

“I don’t want to disparage what we were doing 20 years ago, because I was one of the people doing it. But the toolkits sort of forced you to think two-dimensionally,” he observed.

Maler said well loggers and geoscientists are now more integrated into the upstream process along with engineers and drilling-and-completion specialists. It’s no longer a case of handing off projects and studies down the line from geology to engineering to drilling to completion.

“Now you see much more group effort to figure it all out. And it’s very iterative,” Maler said. “Asset teams are working collaboratively, and thinking of this as a holistic problem.”

fxvtatdcssyrbtasxveebubvzbywfsc

You may also be interested in ...