Using Artificial Intelligence in Fossil Energy R&D

When you hear about artificial intelligence (AI), the first technologies that probably come to mind are voice-activated assistants like Alexa or Google Home, self-driving cars, and drone delivery programs. If you think these innovations are exciting, then we can’t wait to share how we’re using AI here at the Office of Fossil Energy (FE).

Heralded as the next technological revolution, AI makes it possible for machines to learn from experience and perform tasks commonly performed by human beings. The integration of AI into our research and development (R&D) efforts is transforming FE’s ability to analyze massive datasets and solve complex problems.

We have over 60 AI-enabled projects underway. Check out these 5 examples that show how AI R&D is utilizing and protecting our Nation’s vast fossil energy resources:

1. Smart Robots that Inspect and Repair Power Plant Boilers

Power plant boilers are the most important part of a power plant, but they are difficult and time-consuming for human operators to inspect and repair. To reduce risks and shorten maintenance and unplanned outages, FE is developing AI-enabled robots that can perform real-time, non-destructive inspection of boiler furnace walls. If they find a crack, they can operate repair devices to make an immediate repair, while using AI to enable smart data analysis and autonomy.

2. Drone-Mounted, Smart Methane Emissions Detection Systems

Methane emissions are of increasing concern to the oil and natural gas industry. They represent lost product and are a recognized greenhouse gas. To help reduce these emissions, a smart methane emissions detection system is being developed. It will detect methane leaks by pairing passive optical sensing data with AI algorithms. The system will also be mounted onto a drone, which will enable a more precise leak detection method.

3. Models that Predict Oil and Gas Well Productivity After Hydraulic Fracturing

The ability to design a more effective hydraulic fracturing program for a particular basin and predict how much oil and gas can be extracted from it is vital. However, executing a better design and obtaining accurate predictions is difficult because the process occurs underground. That is why AI is being used to make sense of multiple data sets and predict a well’s performance prior to drilling. These models will help energy producers optimize oil and gas stimulation and production and reduce environmental waste.

4. SMART-CS Initiative

Today’s technology can securely store captured carbon dioxide deep in the subsurface of the ground, but slow data processing can result in operational inefficiencies. To meet this challenge, FE developed a Science-Informed Machine Learning to Accelerate Real Time Decisions (SMART-CS) initiative. Using science-based machine learning and AI, this initiative will enable better reservoir management through more rapid decision making. It will develop real-time visualization, forecasting capabilities, and virtual learning environments. As a result, the SMART-CS initiative will help stakeholders and regulators overcome costly inefficiencies while increasing their confidence that the geologic carbon storage is secure.

5. Computers Dedicated to Fossil Energy Research

The Joule 2.0 supercomputer and the WATT computer, both housed at the National Energy Technology Laboratory (NETL), help accelerate the development of innovative, cost-effective technologies to ensure affordable, reliable energy for all Americans.

  • Joule 2.0 supercomputer

Joule 2.0 allows researchers to model energy technologies, simulate challenging phenomena, and solve sophisticated problems using AI and other computational tools. A talented mathematician working 40 hours a week for 50 weeks per year would take about 55.9 billion years to do what Joule 2.0 can do in one second.

  • WATT computer

At the heart of NETL’s Center for Artificial Intelligence and Machine Learning is the WATT computer. It is optimized to rapidly ingest the enormous amounts of data required for ‘deep learning’ AI. To get a feel for how fast the WATT computer can ingest data, if 532 miles of shelves at the Library of Congress were stacked end-to-end, the computer would be able to read the shelves at 15,700 miles per hour, which is around 4.4 miles per second.

These are just a few of the many ways we’re using AI in our R&D efforts. Our successes are made possible through our collaboration with the U.S. Department of Energy’s Artificial Intelligence and Technology Office, industry, academia, and NETL. To learn more about the Office of Fossil Energy and our projects, visit