Challenge
Delivering cost-effective energy from the Earth’s subsurface entails the use of heterogeneous reservoirs, dominated by fractures. Tools capable of predicting reservoir behavior and the interactions of complex biogeochemical, mechanical, and hydrologic processes are critical to the development of innovative, cost-effective extraction of subsurface energy sources, including unconventional oil and gas, geothermal, and coal bed methane.
AI Solution
Developing AI capable of reasoning under extreme uncertainty, integrating heterogeneous data types (i.e., seismic, geochemical, biological, hydrologic), and building predictive models of systems that cannot be directly observed has broad applicability to any domain requiring inference from indirect evidence. For subsurface science, AI that connects molecular-scale mechanisms to field-scale resource availability will transform the field from descriptive to predictive. DOE’s vast biological, geochemical, geomechanical, and hydrologic data sets can be combined with high performance modeling capabilities, laboratory analytics, and a suite of AI technologies, including surrogate models, physics-informed machine learning, and digital twins to enhance engineering evaluation and control of the subsurface during characterization, drilling, stimulation, and production.
Justification
DOE’s laboratory complex includes a vast array of data, modeling, and analytical capabilities (e.g., Science-informed Machine Learning for Accelerating Real-Time Decisions in Subsurface Applications (SMART) Initiative suite of tools and machine learning-based surrogate models; National Energy Technology Laboratory’s Energy Data eXchange (EDX) Discover; LBNL’s Transport Of Unsaturated Groundwater and Heat (TOUGH) simulation suite; Los Alamos National Laboratory’s (LLNL) GeoDTi design tool; LLNL’s GEOS software; and PFLOTRAN and ATS models), that supports extensive research and operational activities in subsurface environments, and are uniquely positioned to apply AI to accelerate development of models that emulate microbial, mineral, and fluid interactions across molecular to field scales. Efforts will assemble, train, and analyze information from vast data libraries, experiments, and operational sensors, to produce subsurface transport models and digital twins to enhance reservoir characterization, enable real-time reservoir simulation, and address uncertainty to increase hydrocarbon resource recovery and unlock geothermal resources.
National Impact
Understanding subsurface complexities and stored energy sources is key to securing the nation’s energy future and will reduce the costs of recovery for oil, gas, and heat, reduce costs of power to U.S. consumers, re-shore manufacturing, and enhance U.S. competitiveness.
Related Challenges
This was last updated .