Spotlight: Stone Ridge Technology Accelerates Reservoir Simulation Workflows with NVIDIA PhysicsNeMo on AWS

Risk and uncertainty inherent in energy exploration include unknown geological parameters, variations in fluid and rock properties, boundary conditions…

NVIDIA logo
NVIDIA
•
Dmitriy Tishechkin
•7 min read•advanced•
--
•View Original
AWSPython

Overview

The article discusses how Stone Ridge Technology accelerates reservoir simulation workflows using NVIDIA PhysicsNeMo on AWS. It highlights the integration of machine learning techniques with high-performance computing to generate full-field proxy models that are significantly faster than traditional simulations while maintaining accuracy.

What You'll Learn

1

How to generate full-field proxy models using NVIDIA PhysicsNeMo

2

Why using machine learning techniques can enhance reservoir simulation accuracy

3

How to implement reservoir simulation workflows on AWS

Prerequisites & Requirements

  • Understanding of reservoir engineering concepts
  • Familiarity with AWS services like S3 and EC2(optional)

Key Questions Answered

How does the integration of NVIDIA PhysicsNeMo improve reservoir simulations?
The integration of NVIDIA PhysicsNeMo allows for the generation of full-field proxy models that are 10x-100x faster than traditional forward simulations while providing reasonably accurate results. This enhances the efficiency of reservoir simulations and enables rapid evaluations of various scenarios.
What are the main components of the workflow for generating proxy models?
The workflow involves several key steps: ensemble generation, forward modeling with ECHELON, ML data preprocessing, training and validation using NVIDIA PhysicsNeMo, and building the proxy model. This structured approach ensures that the generated models are both accurate and efficient.
What challenges does the full-field proxy model address in reservoir engineering?
The full-field proxy model addresses challenges such as uncertainty quantification and field optimization problems. It allows for the rapid evaluation of scenarios that would otherwise require extensive computational resources and time.

Key Statistics & Figures

Speed improvement of proxy models
10x-100x
Proxy models generated using NVIDIA PhysicsNeMo are significantly faster than traditional forward simulations.
Number of wells in the Norne field model
35
The Norne field model includes 22 producers and 9 injectors, showcasing the complexity of the geological setup.

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Machine Learning Framework
Nvidia Physicsnemo
Used for building, training, and fine-tuning physics-ML models for proxy generation.
Cloud Computing
AWS
Provides scalable resources for running simulations and storing data.
Reservoir Simulator
Echelon
Serves as a full-physics simulator for generating training data and validating proxy models.

Key Actionable Insights

1
Implementing NVIDIA PhysicsNeMo can significantly reduce simulation times in reservoir modeling.
By leveraging the capabilities of NVIDIA GPUs and machine learning, engineers can achieve faster results, allowing for more iterations and optimizations in their workflows.
2
Utilizing AWS for reservoir simulations provides scalability and flexibility.
AWS services like S3 for data storage and EC2 for compute resources enable teams to handle large datasets and complex simulations without the need for extensive on-premises infrastructure.
3
Combining traditional simulation methods with machine learning can enhance accuracy.
The use of full-field proxy models allows for a more nuanced understanding of reservoir behavior, improving decision-making in exploration and production.

Common Pitfalls

1
Assuming that traditional simulation methods are sufficient for all reservoir modeling tasks.
Many modern challenges in reservoir engineering require the speed and flexibility that machine learning and cloud computing provide, which traditional methods may not offer.

Related Concepts

Machine Learning In Engineering
High-performance Computing
Reservoir Engineering Techniques