Skip to content

fix: reject demand and solar forecasts above 100 MWh per step - #325

Merged
andig merged 1 commit into
mainfrom
fix/reject-absurd-forecasts
Nov 9, 2026
Merged

andig merged 1 commit into
mainfrom
fix/reject-absurd-forecasts

Conversation

@andig

@andig andig commented Nov 9, 2026

Copy link
Copy Markdown
Member

Follows #321.

About 90 requests a day still end Not Solved on revision 63. All come from one installation whose household profile carries 1e9 to 1.5e16 Wh in nearly every step, repeating every 96 slots. Its grid has no import limit, so the model is feasible, CBC does not prove the probe infeasible and #321 does not apply. The split hands it to HiGHS, which returns no solution for such coefficients.

  • gt and ft now carry maximum: 1e8 Wh per step in the request schema, so such a request is a 400 naming every offending step (time_series.gt.N), like the other field checks
  • 1e8 Wh is 100 MW over an hour step. On hybrid a single step of 1e8 still solves Optimal; plain CBC already fails above 1e6 per step, which this does not address
  • the 32 Not Solved requests from the slow log all replay as 400
  • the fix: report an infeasible probe instead of handing it to the split #321 test used 1e36 Wh to reach an infeasible probe; that is now rejected first, so it simulates the infeasible solve like the dump test does

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@andig
andig merged commit a7a45da into main Nov 9, 2026
1 check passed
@andig
andig deleted the fix/reject-absurd-forecasts branch November 9, 2026 16:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant