Month: January 2020

Frankenstein’s ROMster: Reduced-order modeling in geologic CO2 sequestration

Reduced-order models (ROMs) are a widely used and powerful approach to reducing the complexity of predictive physics-based numerical simulations for a wide range of applications, including electronics and fluid mechanics such as geologic CO2 sequestration (GCS). Note ROMs are also called response surface models, proxy models, meta-models or surrogate models in different fields. ROMs are critical for optimization, sensitivity analysis, model calibration and uncertainty quantification where full-order models cannot be feasibly executed many times. Los Alamos researchers have utilized machine learning based ROMs to address a lot of critical issues in GCS for decade, e.g., data-worth analysis in monitoring design, evaluation of CO2/brine leakage, assessment of CO2 storage capacity.

However, all the ROMs developed in the previous work were based on traditional approach where a single ROM for each simulated response (e.g., CO2 injection rates, CO2/brine leakage rates) is generated based on a set of training simulations. Most recently, SimCCS researchers demonstrated that a single ROM can display excellent overall predictive statistics, but have predictions that dramatically and unacceptably deviate from simulator responses especially when the response variable has a large range (i.e., vary over multiple orders of magnitude). For example, they showed that a traditional statistically-high-performing GCS ROM (coefficient of determination R2 of 0.99) can have average absolute relative errors of over 200%. To address this, they proposed a new and novel approach where a set of sub-ROMs are generated to overcome the potential pitfalls in traditional single ROM development. The effectiveness of the proposed approach—the ROMster framework—is demonstrated using a case study of predicted CO2 injection rates for GCS. They found their approach is a robust and general framework for ROM development, reducing the average “error” from 200% to only 4% in their case study. It is believed that many researchers in GCS, hydrology, petroleum and other communities have not realized the potential pitfall in ROM generation and thus the ROMster framework will provide a powerful approach for future ROM development.

Link to the paper: