Funding push for new open-source MILP solver HiGHS
81 views
Skip to first unread message
Tom Brown
Apr 1, 2022, 7:40:01 AM4/1/22
to PyPSA mailing list
Dear PyPSA users,
You may have seen in PyPSA Version 0.19.0 that we added support for the
new open-source MILP solver HiGHS in an effort led by Max Parzen.
We are now organising a funding push to take HiGHS to the next level and
bring it up to the speed of commercial solvers like Gurobi and CPLEX:
https://pypsa-meets-africa.github.io/highs
Please donate generously and tap your donor networks for bigger sums!
We're aiming to raise USD 114,000 in the first year for the developer
team at the University of Edinburgh led by Julian Hall, with more to
follow. There is a funky PDF for sharing with potential donors.
I'm not sure this crowd needs the hard-sell on this, but here we go:
As a community we have made lots of progress in developing open energy
models (of which there may now be too many...) and some progress in
opening up important energy datasets. But the missing piece has always
been solvers. Academics can get licences for commercial solvers, but
NGOs, non-NGO interest groups, companies, ministries and even academics
running interactive websites cannot afford licences running up to USD
45,000 a year. Unfortunately established open solvers like GLPK and cbc
don't cut it for large energy system problems. Beyond cost, there is
obviously a big benefit for the optimisation community to have insight
into how the code works and develop it further.
HiGHS, led by a research team around Julian Hall at Uni Edinburgh, has
entered the scene and stolen the show with performance that gives Gurobi
a run for its money:
https://forum.openmod.org/t/open-source-highs-solver-performance-boost-for-energy-system-models/2922
Further Mittelmann benchmarks:
https://mattmilten.github.io/mittelmann-plots/
Based on discussions with the developer team, they reckon there is some
lowish-hanging fruit to boost the performance of HiGHS for energy system
problems. In response, Max Parzen has led this effort to coordinate a
funding push, with some help from Julian Hall, Jesse Jenkins and me, as
well as from other members of the modelling community who have offered
"impact statements".
Thanks for helping us boost this effort - the whole energy community
will benefit from performant open source solvers, and probably many
other communities doing optimisation too!
Best wishes,
Tom
--
Tom Brown (he/him)
Professor of Digital Transformation in Energy Systems
Institute of Energy Technology
Technische Universität Berlin
Group website: https://tub-ensys.github.io/
Personal website: https://nworbmot.org/
Visitor Address:
Einsteinufer 25 (TA 8)
10587 Berlin
You may have seen in PyPSA Version 0.19.0 that we added support for the
new open-source MILP solver HiGHS in an effort led by Max Parzen.
We are now organising a funding push to take HiGHS to the next level and
bring it up to the speed of commercial solvers like Gurobi and CPLEX:
https://pypsa-meets-africa.github.io/highs
Please donate generously and tap your donor networks for bigger sums!
We're aiming to raise USD 114,000 in the first year for the developer
team at the University of Edinburgh led by Julian Hall, with more to
follow. There is a funky PDF for sharing with potential donors.
I'm not sure this crowd needs the hard-sell on this, but here we go:
As a community we have made lots of progress in developing open energy
models (of which there may now be too many...) and some progress in
opening up important energy datasets. But the missing piece has always
been solvers. Academics can get licences for commercial solvers, but
NGOs, non-NGO interest groups, companies, ministries and even academics
running interactive websites cannot afford licences running up to USD
45,000 a year. Unfortunately established open solvers like GLPK and cbc
don't cut it for large energy system problems. Beyond cost, there is
obviously a big benefit for the optimisation community to have insight
into how the code works and develop it further.
HiGHS, led by a research team around Julian Hall at Uni Edinburgh, has
entered the scene and stolen the show with performance that gives Gurobi
a run for its money:
https://forum.openmod.org/t/open-source-highs-solver-performance-boost-for-energy-system-models/2922
Further Mittelmann benchmarks:
https://mattmilten.github.io/mittelmann-plots/
Based on discussions with the developer team, they reckon there is some
lowish-hanging fruit to boost the performance of HiGHS for energy system
problems. In response, Max Parzen has led this effort to coordinate a
funding push, with some help from Julian Hall, Jesse Jenkins and me, as
well as from other members of the modelling community who have offered
"impact statements".
Thanks for helping us boost this effort - the whole energy community
will benefit from performant open source solvers, and probably many
other communities doing optimisation too!
Best wishes,
Tom
--
Tom Brown (he/him)
Professor of Digital Transformation in Energy Systems
Institute of Energy Technology
Technische Universität Berlin
Group website: https://tub-ensys.github.io/
Personal website: https://nworbmot.org/
Visitor Address:
Einsteinufer 25 (TA 8)
10587 Berlin
Max Parzen
Jul 21, 2022, 4:24:19 AM7/21/22
to pypsa
Big news, HiGHS got 75000€ funding. Thanks all! Hope more is coming.
- #Twitter thread on this news by Tom Brown: https://twitter.com/nworbmot/status/1549796252107345920?s=20&t=wpYfP-Q05Zs3VXuHQiLJhw
- #miniproposal version with DOI: "Optimization solvers: the missing link for a fully open-source energy system modelling ecosystem" https://zenodo.org/record/6534004#.YtkJNOzMKEA
- #openmod, the place where everything started with a benchmark: https://forum.openmod.org/t/open-source-highs-solver-performance-boost-for-energy-system-models/2922/10
Best wishes,
Max
Reply all
Reply to author
Forward
0 new messages