As the world faces the urgent need to combat climate change, carbon capture and storage (CCS) has emerged as a crucial technology for achieving net-zero…
Overview
Shell, in collaboration with NVIDIA, has developed a machine learning model using Fourier neural operators to enhance the efficiency of carbon capture and storage (CCS) site screening. This innovative approach achieves a computational speedup of 100,000 times compared to traditional methods while maintaining high accuracy in predicting CO2 plume migration and pressure buildup.
What You'll Learn
1
How to leverage AI-based surrogate models for CO2 storage modeling
2
Why machine learning can significantly speed up CCS site screening processes
3
How to apply Fourier neural operators in subsurface modeling
Prerequisites & Requirements
- Understanding of carbon capture and storage (CCS) concepts
- Familiarity with machine learning principles(optional)
- Access to NVIDIA PhysicsNeMo framework(optional)
Key Questions Answered
How does Shell use NVIDIA PhysicsNeMo for CO2 storage modeling?
Shell utilizes NVIDIA PhysicsNeMo to develop machine learning models that enhance the efficiency of carbon capture and storage (CCS) site screening. This collaboration allows for rapid high-resolution modeling of CO2 storage, significantly reducing computational costs while maintaining accuracy in predicting plume migration and pressure buildup.
What are the benefits of using AI-based surrogate models in CCS?
AI-based surrogate models reduce computational costs associated with traditional numerical models, enabling the assessment of tens of thousands of injection configurations. This capability facilitates rapid screening of potential storage sites, ultimately contributing to more informed decision-making in CCS applications.
What is the computational speedup achieved with the new model?
The newly developed model achieves a computational speedup of O(10^5) compared to traditional numerical simulators for CO2 flow fields, allowing for efficient site assessments with minimal sacrifice in prediction accuracy.
What challenges does CO2 migration pose in CCS?
CO2 migration poses significant challenges, including the risk of environmental escape from the reservoir and the need to manage pressure buildup to prevent geological cracks and seismic hazards. Accurate modeling is crucial for ensuring safe and effective storage.
Key Statistics & Figures
Computational speedup
100,000 times
Achieved by the machine learning model compared to traditional numerical simulators.
Model training duration
1,000 years
The model is trained on a dataset simulating CO2 behavior over this extensive period.
Technologies & Tools
Framework
Nvidia Physicsnemo
Used for building, training, and fine-tuning Physics-ML models.
Algorithm
Fourier Neural Operators (fnos)
Employed to emulate CO2 behavior in subsurface reservoirs.
Key Actionable Insights
1Integrating machine learning models into CCS workflows can drastically improve efficiency and accuracy.By adopting AI-based approaches, organizations can enhance their ability to screen potential CO2 storage sites, leading to faster decision-making and better compliance with environmental regulations.
2Utilizing NVIDIA PhysicsNeMo can simplify the development of complex physics-based models.This open-source framework provides tools for building and training machine learning models, making it accessible for domain scientists and engineers to apply advanced techniques in their projects.
3Regularly assess the accuracy of your models using physics-based metrics.Implementing rigorous validation processes ensures that your predictions remain reliable, especially when dealing with critical environmental applications like CCS.
Common Pitfalls
1
Neglecting the importance of accurate site assessments can lead to environmental risks.
In CCS, improper evaluation of storage sites may result in CO2 leakage, which can have severe ecological consequences. It's crucial to use reliable models and metrics to ensure safe storage.
Related Concepts
Carbon Capture And Storage (ccs)
Machine Learning In Environmental Science
Subsurface Modeling Techniques