CCSNet

Stanford/Caltech’s digital twin for ML-based CCS modeling leverages Nvidia Omniverse and Modulus.

Presenters from Nvidia, Stanford University and Caltech have presented a model (which we recommend reading for its comprehensive illustrations) for the use of digital twins in carbon capture and storage. Currently, some 30 large-scale CCS operations sequester some 40Mt of CO2* per year. Development of this technology is set to ‘grow rapidly’ in the coming decade, ‘but this promising solution has yet to prove that it can be industrialized at an acceptable cost’. Seemingly, a key challenge for keeping CCS solutions economical is the cost of proving the duration and reliability of storage using numerical modeling**. Traditional simulators for carbon sequestration are time-consuming and computationally expensive. Enter Machine learning models that promise ‘similar accuracy and significantly reduced time and costs’.

The post, ‘Accelerating climate change mitigation with machine learning: the case of carbon storage’ describes the use of Nvidia Modulus and Omniverse tools to investigate various geological scenarios and injection patterns. The post describes a ‘physics-informed machine learning’ approach that leverages a Fourier neural operator (FNO) to predict the 3D reservoir behavior. The nested FNO approach is claimed to offer an ‘inference speed that is 700,000x faster than a state-of-the-art numerical solver’. One assessment of uncertainties in capacity estimation and injection designs that would have taken nearly two years with numerical simulators ran in ‘only 2.8 seconds’.

The trained nested FNO is hosted on a public GPU-based web application such that users can construct any random combination of reservoir condition, injection scheme, and permeability field characteristics and obtain instantaneous predictions of gas saturation, pressure buildup and sweep efficiency estimates. The web app promotes equity in CO2 storage project development and knowledge adoption. This especially benefits small to mid-sized developers as well as communities that want an independent evaluation of projects being proposed. High-quality forecasts were previously unattainable for these important players. The ML framework leverages Nvidia’s Modulus and Omniverse platforms. Modulus is a physics-ML framework for developing physics-based, machine-learning models. Modulus takes both data and the governing physics into account to train a neural network that creates an AI surrogate model for digital twins. Omniverse adds a virtual reality based GUI for interactive exploration of digital twins using the surrogate model output from Modulus.

The nested FNO model on is available on Git. Read the original paper here on ‘Accelerating carbon capture and storage modeling using Fourier neural operators’.

* World carbon emissions in 2022 were estimated at some 37 gigatons.

** Computer modeling may not be a major cost compared with the astronomical sums involved in engineering CCS!

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.