ExaGOTM is a package for solving large-scale power grid optimization problems on parallel and distributed architectures, particularly targeted for exascale machines with heteregenous architectures (GPU). Combinations of stochastic, contingency-constrained, multiperiod ACOPF problems can be solved with ExaGO. The package is written in C/C++ with python bindings available for python-based applications. An overview of the package is given on this page. For extended information, including the modeling details and formulations, see the ExaGO manual.
ExaGOTM includes the following applications for solving different power grid optimization problems:
- OPFLOW solves an AC optimal power flow either on CPU and GPU
- TCOPFLOW solves a multi-period optimal power flow
- SCOPFLOW solves a security-constrained (contingency-constrained) optimal power. Both single-period and multi-period problems can be solved.
- SOPFLOW solves a stochastic optimal power flow with (optional) security constraints for single and multiple periods.
ExaGOTM applications are interfaced with the following optimization solver packaages:
- Ipopt is a popular optimization package for solving nonlinear optimization problems that uses an interior-point algorithm.
- HiOp is a HPC package for optimization. ExaGO interfaces with two of its solvers -- a mixed sparse-dense interior-point solver (NewtonMDS) and a sparse interior-point solver (HiOPSparse). NewtonMDS allows execution of the optimization either on CPU and GPU. The sparse HiOp solver is currently supported on CPU only.
Note that not all applications can utilize all solvers yet. The following table lists the solver-application compatibility.
| Solver | OPFLOW | TCOPFLOW | SCOPLOW | SOPFLOW |
|---|---|---|---|---|
| Ipopt | Y | Y | Y | Y |
| HiOp | Y | Y | Y |
Additionally, note that SCOPFLOW and SOPFLOW with HiOp solver use Ipopt to solve a portion of the problem (base problem). So one must also configure with Ipopt when using HiOp solver for these applications.
Details installation instructions are given at INSTALL.md for information on acquiring, building and installing ExaGO.
If you are a developer with access to the project, we also provide public binaries that are generated through our GitHub actions workflows documented in README.md, and with documentation about usage in the packages section of our repository. Check out a short (< 60s demo) of pulling down a version of ExaGO:
You can view the following helpful documentation sources:
- test_add.md markdown file for information on adding tets (outdated)
- README.md for our bash / spack buildsystem used in GitHub/GitLab CI/CD
- README.md for our spack specific build scripts that support CI tcl modules on HPC target platforms
- README.md for our devcontianer configuration information (codespace support coming soon)
- exago_policy_compatiblility for xSDK compatibility guidelines, and ways to enforce compliance
- python_bindings.md for documentation about or Python bindings
- README.md for information about profiling ExaGO with spack
- README.md for details about our GitHub actions
- README.frontier.md for ORNL's Frontier specific configuration
Our ChatGrid frontend deployed with React, PSQL and LangChain has documentation in README.md as well as a pdf README.pdf in the viz subdirectory. Several of our tutorials install this through commands in Jupyter Notebooks as well.
Instructions for executing the different ExaGOTM applications is given below.
We also provide our user manual as a pdf manual.pdf.
- If you are using a devcontainer with VSCode, the following tutorials are provided:
- tutorial.ipynb for basic configuration infromation and I/O
- mpi4py-tutorial.ipynb for mpi4py pointers and best practices
- viz-tutorial.ipynb for spinning up our frontend visualization with ChatGrid integration
- Otherwise, you can check out our more in depth application tutorials in the
tutorialssubdirectory:- demo1.ipynb run OPFLOW, SCOPFLOW and visualize your output
- demo2.ipynb run SOPFLOW on many ranks using MPI, and visualize outpu
- TODO - add fixes from
mpi4pydevcontainer example into this notebook to show working MPI workflow
- TODO - add fixes from
Each application has a different set of options that are described in depth in
the usage notes. These options can be passed optionally through an options file
(-optionsfile <option_file>), or directly on the command line.
Since options may be specified in more than one location (on the command line,
and through an options file), it is worth noting that the option specified on
the command line supersede those in the options file. For example, if
opflowoptions options file set the network file via the option
-netfile case9mod.m, the following behavior occurs:
# This uses case9mod.m
./opflow -optionsfile opflowoptions
# This uses case118.m
./opflow -netfile case118.m -options_file opflowoptionsExaGO has an experimental visualization to display the results of OPFLOW
application on a map. See the visualization README for more
information.
Please see the developer guidelines before attempting to contribute. Feel free to raise an issue or contact the team if the guidelines are ambiguous or you have a particular question.
ExaGOTM was designed and implemented by Shrirang Abhyankar and has received significant from Slaven Peles, Nicholson Koukpaizan, Maksudul Alam, Asher Mancinelli, Cameron Rutherford, Joshua Hambrick, Philip Fackler, Eve Tsybina, Bruce Palmer, Jaelyn Litzinger, William Perkins, Sayef Azad Sakin, Joseph Macam, and Ryan Danehy.
ExaGOTM has been originally developed as a part of ExaSGD project under the Exascale computing project.
Copyright © 2026, UT Battelle.
ExaGOTM (v2) is a free software distributed under a BSD-style license. See the LICENSE and NOTICE files for details. All new contributions to ExaGOTM must be made under the same licensing terms.
Please Note If you are using ExaGOTM with any third party libraries linked in (e.g., CoinHSL), be sure to review the respective license of the package as that license may have more restrictive terms than the ExaGOTM license.