Design and Evaluation of High-Efficiency Boilers for Energy Production Using a Hierarchical V/UQ Approach

PI Name: 
Martin Berzins
Institution: 
The University of Utah
Allocation Program: 
Aurora ESP
Year: 
2017
Research Domain: 
Chemistry

For this project, researchers will simulate and evaluate the design of a next-generation 500MW Advanced Ultra Super-Critical (AUSC) coal boiler. Application of a novel computer science approach in conjunction with a hierarchical validation and uncertainty quantification methodology (V/UQ) should result in a simulation tool that can predict the thermal performance of the boiler with uncertainty bounds as constrained by observed data across multiple physical scales. The simulation data and resulting simulation tool have the potential of impacting boiler design by minimizing capital investment, a 50% increase in boiler efficiency, and an associated 50% reduction of CO2 greenhouse gases.

This work will be undertaken by the University of Utah Carbon Capture Multi-Disciplinary Simulation Center (CCMSC), one of three large DOE NNSA PSAAP II Centers funded from 2014 to 2019. The CCMSC aims to use simulations at petascale and beyond to aid the design of this and future generations of clean boilers. The center is comprised of multidisciplinary scientists and engineers from the universities of Utah, California Berkeley, and Brigham Young. The center enjoys a tight collaboration with GE-Alstom, a world leader in boiler design and manufacturing. This has resulted in data-sharing agreements that strengthen the industrial applicability of the research under CCMSC and enables the center’s goal of using simulation to accelerate deployment of new energy technology in industry.

A major component of the CCMSC program is the development and deployment of simulation software for the problem of interest, namely the Arches component in the Uintah computational framework. Arches is a multi-scale, large-eddy simulation tool, solving coupled particle and gas combustion in low-Mach systems with radiation effects. The tool relies on sub-modeling for unresolved physics and uses a linear solve for solving low-mach number derived equations. Uintah uses a novel asynchronous task-based methodology that supports dynamic out-of-order task execution as well as the use of MPI, MPI-threads, with or without accelerators/co-processors. The portability of Uintah to present and future architectures relies on a mixture of dynamic asynchronous task-based execution coupled to the Kokkos portability library.

This ensures both portability and nodal performance without changes to the underlying physics. Scalable I/O performance using PIDX to 500K cores along with in-situ visualization with VisIt are important indicators of future approaches. The benefits of this approach were shown in early 2016 when the full physics application weak scaled to 500K Mira cores on a first attempt, which included a scalable linear solver and I/O as part of the team’s 2016 INCITE allocation.

Impact

The continued development of sustainable, clean, and economic energy is essential to address today’s energy challenges. Through HPC and hierarchical V/UQ simulation, the proposed boiler design can be transformational in accelerating the deployment of high efficiency coal boilers, and meeting our energy challenges.