The Atikokan Digital Twin is the capstone project of the Carbon Capture Multidisciplinary Simulation Center (CCMSC). CCMSC was established in April of 2014 at The University of Utah by the United State Department of Energy (DOE), National Nuclear Security Administration (NNSA) for the purpose of developing and demonstrating the use of formal uncertainty quantification (uq) methodologies in conjunction with scalable and portable high performance computing (hpc) strategies for solving large practical problems. To accomplish this mission, we developed a multi-physics, large-eddy simulations (LES) code (Arches) to run at scale on world-class computational resources made available to us by NNSA. The application selected by the center, was the demonstration of positive societal impact of hpc with uq for the deployment of low-cost, low-carbon energy solutions for power generation. To guide our application we partnered with two industrial collaborators, General Electric (GE Power) and Ontario Power Generation (OPG). Faculty, staff and students from the University of California, Berkely, and Brigham Young University participated with those from the University of Utah. These academics were comprised of 78 engineers and scientist working together in three teams to complete the center mission: the computer science team, the physics team, and the uq team. The federal project was completed in January 2021 with a total financial contribution of \$28 million (\$23.2million-federal, \$5.2million-university, industrial partner contributions were in-kind), at which time CCMSC was dissolved.
The CCMSC industrial partners and their applications provided purpose and focus to the methodologies developed in the center. With GE Power, our objective was to demonstrate the advantages of hpc with hierarchical uq in design decisions by predicting the heat flux profile to a validated level of uncertainty for a full-scale, pulverized solid-fuel (coal) thermal power generation boiler. Our capstone project was completed in partnership with OPG, where we deployed all the methodologies of the center to demonstrate dynamic, online artificial intelligence (ai) for operating a biomass-fired power generation boiler. The Atikokan Digital Twin starts with a large suite of our validated, multi-physics, LES simulations run on hpc resources and selected from a design of experiments covering the full potential operational space for the power boiler. This suite of Arches simulations is then abstracted into surrogate models for all quantities of interest. Then the digital twin ai uses our science-based Bayesian machine learning (ml) methods to combine these surrogate models with online power-plant measurements to produce real-time (3-5 minute updates) operational set points for continuous optimization of the biomass boiler in the presence of uncertainty.
A four and a half minute video illustrating the capabilities of the Arches simulator.
A fifteen minute conference-style presentation on the Atikokan Digital Twin.
A video presentation providing background and an overview of this digital twin.
(note: use ‖ ► and the arrow keys to control the pace of the presentation)
The CCMSC Final Technical Report to DOE-NNSA describing this digital twin.
(note: section 8 contains a list of 250 CCMSC publications w/ DOI links)
The digital twin provides both online optimization and offline interrogation of 'what-if' scenarios.
Links are provided below to the demonstration of these tools, along with access to a tool
for visualizing results from the 420 Arches simulations used to build the digital twin ai:
is a demonstration of the online optimization for operating conditions from December 2019.
............................................................................................................................debug version
is a tool for presenting the predicted boiler outputs from user selected operating conditions.
............................................................................................................................debug version
is a tool for visualizing results from the 420 Arches LES simulations of the Atikokan boiler.
............................................................................................................................debug version