openego / eDisGo / 42238117639

87%

push

github

Write the OPF curtailment back onto the full grid in spatial_restore (#855)

* Write the OPF curtailment back onto the full grid in spatial_restore

opf_version 3 and 4 curtail the solar and wind generators and write the
curtailed feed-in onto the grid they optimised. With spatial complexity
reduction this is the reduced grid, and apply_reduced_results_to_full_grid
only wrote the flexibilities back, so the pipeline reinforced the full grid
for the uncurtailed feed-in.

apply_reduced_results_to_full_grid now always writes back all generators,
by name or, for merged generators, split in proportion to each member's
uncurtailed feed-in. Generators the OPF did not change get their unchanged
series back, as a merged series is the sum of its members.

Fixes #850

* Test the physical invariants of the curtailment write-back

test_curtailment_reaches_the_full_grid now curtails by a different
fraction per generator and time step and checks what the OPF result means
physically: each reduced generator's dispatch equals the sum of its
full-grid members, every generator stays between 0 and its potential,
reactive power keeps its Q/P ratio, and uncurtailed generators keep their
series.

test_opf_without_curtailment_request_keeps_the_feed_in runs the real OPF
(opf_version 3, runonlinux) on the reduced test grid without a curtailment
request and checks that the full grid keeps its feed-in, up to the
solver's tolerance.

* Address review: Ipopt in the OPF test, split check, stale texts

- test_opf_without_curtailment_request_keeps_the_feed_in solves with
  method="nc" (Ipopt), as the GitHub runners have no Gurobi license.
- test_curtailment_reaches_the_full_grid scales the worst-case feed-in per
  generator before the reduction and checks each member's share, so a split
  by p_nom no longer passes.
- apply_reduced_results_to_full_grid documents the limits of the generator
  write-back; the _disaggregate docstring and parameter names describe the
  weights it is given.
... (continued)

9 of 9 new or added lines in 1 file covered. (100.0%)

10212 of 11759 relevant lines covered (86.84%)

0.87 hits per line