Deploying High Performance Computing Applications of AI/ML in the Energy Sector

Recorded On: 09/16/2021

Recent advances in autonomous learning and artificial intelligence algorithms have allowed for breakthrough calculations in a number of fields – particularly excelling at efficient data-driven forward modeling and solving computationally intensive inverse modeling problems. We present an overview of state-of-the-art reinforcement learning strategies and GPU-accelerated computational frameworks. We will briefly cover applications of these methods to model problems including mathematical puzzles, mining optimization and path planning, followed by a deep dive into the industrial problem of well placement optimization. Here, reservoir simulations are coupled to a Soft-Actor-Critic reinforcement learning technique that predicts optimal locations and temporal steps to drill wells to maximize the Net Present Value of cumulative production. The optimal results obtained will be shared for a benchmark reservoir model. Subsequently, tangible breakthroughs enabled by deep reinforcement learning over conventional numerical strategies, as well as the future of HPC and AI in the energy sector will conclude the presentation.

This webinar is categorized under the Data Science and Engineering Analytics technical discipline.

All content contained within this webinar is copyrighted by Vidyasagar and Nefeli Moridis and its use and/or reproduction outside the portal requires express permission from Vidyasagar and Nefeli Moridis.

Vidyasagar

Senior Machine Learning Scientist, Beyond Limits

Vidyasagar is a Senior ML Scientist at Beyond Limits. He develops novel reinforcement learning strategies to tackle enterprise-grade problems and accelerate digital transformation across industries. Previously, he worked on full-GPU porous media flow simulations and HPC. Vidyasagar obtained his PhD from Caltech and has authored numerous journal and conference publications on scientific computing.

Nefeli Moridis

Developer Relationship Manager, Subsurface, NVIDIA

Nefeli Moridis is the developer relationship manager for subsurface applications at NVIDIA, working in the global energy team. Prior to joining NVIDIA, Nefeli worked in the oil and gas industry as a reservoir engineer and as a consultant, focusing on projects in unconventional reservoirs in the U.S. and internationally. She has published eleven papers based on her research and work, and has presented at SPE conferences worldwide. Her current position with NVIDIA focuses on helping to accelerate subsurface applications, such as reservoir simulation, onto GPUs. She supports the digital transformation and energy transition projects and works on machine learning and AI approaches to reservoir engineering problems. She also works to grow the energy team’s reach in greener energy initiatives, expanding into carbon storage and sequestration projects for the industry’s decarbonization efforts. Nefeli holds Ph.D. and M.Sc. degrees in Petroleum Engineering from Texas A&M University, a M.Sc. from the Insitut Français du Pétrole in Reservoir Engineering, and a B.Sc. from the University of Texas at Austin in Petroleum Engineering.

SPE Webinars are FREE to members courtesy of the

image

Key:

Complete
Failed
Available
Locked
Deploying High Performance Computing Applications of AI/ML in the Energy Sector
09/16/2021 at 10:00 AM (EDT)   |  60 minutes
09/16/2021 at 10:00 AM (EDT)   |  60 minutes
Survey
20 Questions
Certificate
0.10 CEU/1.0 PDH credits  |  Certificate available
0.10 CEU/1.0 PDH credits  |  Certificate available