As a young boy in Taiwan, NETL’s Chung Yan Shih, Ph.D., enjoyed playing with Legos and similar toys.
But while many children tended to place their Legos in large piles, Shih preferred to organize his blocks into small, separate groupings based on size and other characteristics. He found that approach made it possible to piece together buildings and other structures that sprang from his inquisitive mind with assembly-line efficiency.
“I was always interested in new stuff and challenging stuff. I dreamed about finding ways to make processes more efficient,” Shih recalled.
Today, as a senior strategic data scientist, Shih is living that dream.
Using artificial intelligence (AI), machine learning (ML) and big data analytics algorithms, Shih and his NET colleagues are resolving complex research challenges to provide Americans with affordable and reliable energy from the nation’s abundant fossil fuel resources.
One focus area for Shih, who joined NETL in 2010, is using computational analysis to improve efficiency, productivity and environmental safety in the natural gas industry. A current project focuses on predicting the estimated ultimate recovery (EUR) from natural gas wells. EUR is a critical estimation for producers and utilities to assess the natural gas supply to meet the needs of home, business and industry usage. It is also a key element to measure the amount of gas that can be extracted from shale formations.
“Many variables come into play,” Shih said. For instance, wells differ significantly based on their location, length of time they have been in production, their design and the resource extraction technology that is used.
The gas reservoir may also vary in geologic properties, such as the thickness of the shale layer and gamma ray levels — naturally occurring radiation produced by shale and other sedimentary rock. The variabilities make it difficult to predict EUR, as well as understand the relationship between these factors and production.
ML is well-suited to make sense of these complicated datasets. Using algorithms, the engines of ML, Shih looks for patterns in the data to identify key drivers of production.
“At this point, there may still be a lot of unknowns, but if the model or algorithm we are developing shows promising results, it can be refined with more information to give us an even better understanding of what’s going on underground,” Shih said.
With further development, the model can make it possible to determine a well’s EUR before drilling begins. “That’s the beauty of machine learning.” Shih added.
Shih’s work also has applications for hydraulic fracturing for shale gas, which involves the injection of water, sand and small amounts of highly diluted chemicals at high pressure down and across into horizontally drilled wells as far as 10,000 feet below the surface. The pressurized mixture causes the rock layer to crack. These cracks, or fissures, are held open by the sand particles so that natural gas from the shale can flow up the well.
Models Shih and his colleagues developed can be used to provide reliable estimates about how much water and sand should be required in the hydraulic fracturing process.
NETL pulls data from a variety of databases and other sources to review production records and other key drivers of well production. Shih also works with NETL’s highly skilled geologists to obtain information.
Jimmy Thornton, associate director of the Computational Science and Engineering directorate, calls Shih a key contributor in NETL’s efforts to use AI and ML as cross-cutting technology to enhance a broad range of energy-related challenges.
“I think Chung is becoming one of our true stars as we develop this technology,” Thornton said.
For his part, Shih appreciates that support and vote of confidence, although, he pointed out, any discussion about “stars” might be better addressed by his wife, Mei-Yu Wang, Ph.D., a postdoctoral fellow at Carnegie Mellon University who specializes in astrophysics.