OpenStudio Analysis Framework

The OpenStudio® Analysis Framework (OSAF) makes the optimization and parametric analysis of building energy models accessible to architects, engineers, designers, and policymakers.

Open-source and scalable to facilitate wider adoption, this framework has a clearly defined application programming interface (API) that other applications can be built upon. It can run on high-performance computing systems, within cloud infrastructure, and on laptops, and uses a common workflow to enable different classes of algorithms.

Solution To Address Key Challenges

Integrating building energy modeling into the building design and retrofit process has been a long-term goal of many building scientists and practitioners. Ideally, potential improvements to buildings could be quickly evaluated based on their impact on various metrics, such as energy consumption, life cycle cost, peak demand, indoor environmental quality, and greenhouse gas emissions.

Calculating such metrics is time-consuming and computationally intensive. To generate these metrics, a practitioner needs access to a computing platform that is easily configurable as well as computing resources that can scale to the problem being solved. To conduct a thorough analysis in a realistic timeframe, users need access to various algorithms to perform sensitivity analyses, uncertainty quantification, design optimization, and model calibration. These algorithms should be easily configurable and effortlessly interchangeable without having to completely reformulate the problem.

NREL's OSAF provides a solution for these practitioners.


Customizable workflows enable analysis of multiple challenges in the building design and retrofit process, using the same simulation platform. Optimization and calibration algorithms are used to find inputs that minimize an objective function, such as heating or cooling energy, for an optimal design problem. The OSAF can also include the error between simulated and measured energy end uses (e.g., electric or gas consumption) for a calibration problem. These algorithms are easily configurable and effortlessly interchangeable without having to completely reformulate the problem.

Following are the key capabilities of the OSAF:

  • Multi-objective optimization and calibration algorithms
  • Single-objective optimization and calibration algorithms
  • Sensitivity analysis and design of experiment methods
  • Sampling-based uncertainty propagation.


The architecture of OSAF leverages Kubernetes (a system for managing containerized applications) to organize different instances that communicate with each other and various external problem-defining applications through a well-defined API. A typical configuration consists of one computer or compute node that functions as the server node, and several other computers or compute nodes that function only as workers. The server node is responsible for analysis management, simulation queuing (either algorithmic or batch processing), and results management. It is deployable on several cloud platforms utilizing the OpenStudio Helm Charts.

The OSAF simulation process is depicted below and outlines the path of a submitted analysis description file through the Docker Swarm architecture. The process consists of 12 steps:

  1. Description file (JSON) submitted to the server through the API
  2. Problem definition sent to web-background to keep the long running analysis in state
  3. Problem definition sent to the R algorithm
  4. R algorithm determines variables to run and posts to web-background to create a datapoint JSON for each datapoint
  5. Datapoint JSON created from the variable values and queued
  6. Worker gets datapoint JSON from the queue and creates a measure-based OpenStudio Workflow file
  7. OpenStudio Workflow file is run on the worker and generates results, including objective functions if specified
  8. Results and objective functions sent to web (through an HTTP POST request method) over the Docker Swarm network
  9. Objective functions sent to the algorithm for possible iteration back to step 4
  10. JSON results posted to the database
  11. Database results made available to the web front-end
  12. Front-end graphical user interface and results API have access to results through the web service.


Private sector and government organizations have used OSAF successfully, either through their existing algorithmic and visualization capabilities or linking the large-scale simulation results to external visualization and analytic tools. Global partners include government, national laboratories, and industry.

The codebase is open source with an API and command line interface. This framework enables users and private sector organizations to easily integrate scalable analytic tools into their own solutions.


An Open-Source Analysis Framework for Large-Scale Building Energy Modeling, Journal of Building Performance Simulation (2020)