Lead Performers:
-- Lawrence Berkeley National Laboratory – Berkeley, CA
-- Oak Ridge National Laboratory – Oak Ridge, TN
-- National Renewable Energy Laboratory – Golden, CO
-- GARD Analytics – Arlington Heights, IL
-- Florida Solar Energy Center – Cocoa, FL
-- Big Ladder Software – Denver, CO
DOE Total Funding:  $1,500,000
FY18 DOE Funding:  $500,000
Project Term:  October 1, 2018 – September 30, 2021
Funding Type:  Competitive Lab Call 2018
Related Projects:  EnergyPlus, Spawn

Project Objective

EnergyPlus has a broad set of advanced capabilities and can model many new energy-efficiency technologies and strategies. Along with transparency and documentation, these capabilities have driven growth in EnergyPlus acceptance and adoption. One area in which EnergyPlus still conspicuously lags other common building energy management (BEM) engines is execution time. Despite continuous incremental improvements over the past several years, EnergyPlus remains more than an order of magnitude slower than its predecessor DOE-2.

This project will use a holistic approach to improve EnergyPlus’ annual simulation run-time by a factor of 10 relative to the October 2018 9.0.0 release.  The team will collect and develop a suite of models representing diverse use cases, workflows, and levels of complexity, and establish an overall baseline run time of the suite that the projected improvement will be based on. The team will profile the models to identify performance hotspots, and use correlation techniques to identify problematic component models or structures. The team will use multiple approaches to achieve the anticipated speedups:

  • Refactoring: restructuring major routines including surface and zone heat balance, plant loop simulation, and HVAC simulation with new data structures, improved algorithms, and optimized convergence criteria
  • Vectorization: leveraging new data structures and microprocessor vector extensions to achieve fine grain zero-overhead parallelization
  • Parallelization: using multiple cores on modern processors and GPUs to achieve coarse-grain parallelism where possible
  • Caching: Saving intermediate results for reuse in later simulations within the same workflow

As they are tested and proven out, the enhancements will be incrementally integrated into EnergyPlus using the primary feature development process. Industry partners will provide input to the project and participate in performance testing.

Project Impact

Reducing EnergyPlus runtime improves user productivity in conventional design and analysis tasks and supports quick decision making (e.g., design charrette) on key strategies of energy efficiency for new and existing buildings. Speeding up EnergyPlus will enable broader use in parametric runs, optimization of designs and controls, and urban scale modeling (which can require many simulation runs). The 10X performance speed up will enhance EnergyPlus adoption. The related code refactoring also benefits EnergyPlus code modularization and ease of maintenance.


DOE Technology Manager: Amir Roth
Lead Performer: Tianzhen Hong, Lawrence Berkeley National Laboratory