Pull request #1620 (with the follow-up #1641 and #1648) introduces a secondary energy dimension in subsector-specific industry FE demand.
It was designed to fix performance issues, but in fact convergence is better without it (see images below).
This is due to a flat optimum in the new dimension, as it does not make a difference which SE ends up in which subsector.
This PR reverts the main changes in these PRs while keeping some others, like a price change limit between calibration iterations.
Credit for all the performance tests go to @0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q; Thanks for that!
Type of change
(Make sure to delete from the Type-of-change list the items not relevant to your PR)
[ ] Bug fix
[ ] Refactoring
[ ] New feature
[x] Minor change (default scenarios show only small differences)
Purpose of this PR
Pull request #1620 (with the follow-up #1641 and #1648) introduces a secondary energy dimension in subsector-specific industry FE demand. It was designed to fix performance issues, but in fact convergence is better without it (see images below). This is due to a flat optimum in the new dimension, as it does not make a difference which SE ends up in which subsector.
This PR reverts the main changes in these PRs while keeping some others, like a price change limit between calibration iterations.
Credit for all the performance tests go to @0UmfHxcvx5J7JoaOhFSs5mncnisTJJ6q; Thanks for that!
Type of change
(Make sure to delete from the Type-of-change list the items not relevant to your PR)
Checklist:
remind2
where it was neededforbiddenColumnNames
in readCheckScenarioConfig.R in case the PR leads to deprecated switchesFAIL 0
in the output ofmake test
)CHANGELOG.md
has been updated correctlyFurther information (optional):
/p/tmp/pehl/Remind_selectively_revert_1620/output/
H12:
EU21:
![image](https://github.com/remindmodel/remind/assets/44020564/1c2d0ca0-b880-4814-9d16-6c679c3d39b3)