test/test_native_dcp_acp_models.jl:66, testset "native DCPNetworkModel
solves under network reduction (c_sys14)", carries:
# TODO: fails with HiGHS QP solver, why?
model = DecisionModel(template, sys; optimizer = ipopt_optimizer)
The model is NetworkModel(DCPNetworkModel; reduce_radial_branches =
true, reduce_degree_two_branches = true) on c_sys14 with a thermal
dispatch template — linear network physics with quadratic costs, i.e.
a convex QP that HiGHS supports. A failure there points at the model,
not at solver capability: a bad coefficient or degenerate constraint
from the reduction path is more likely than a HiGHS limitation.
The workaround also removed the only HiGHS coverage of the reduction
path, so nothing now exercises reductions against an LP/QP solver.
Acceptance: root cause identified, and either the testset restored to
HiGHS_optimizer or the comment replaced with a specific explanation of
why Ipopt is required.
Introduced in #225.
test/test_native_dcp_acp_models.jl:66, testset "native DCPNetworkModel
solves under network reduction (c_sys14)", carries:
The model is NetworkModel(DCPNetworkModel; reduce_radial_branches =
true, reduce_degree_two_branches = true) on c_sys14 with a thermal
dispatch template — linear network physics with quadratic costs, i.e.
a convex QP that HiGHS supports. A failure there points at the model,
not at solver capability: a bad coefficient or degenerate constraint
from the reduction path is more likely than a HiGHS limitation.
The workaround also removed the only HiGHS coverage of the reduction
path, so nothing now exercises reductions against an LP/QP solver.
Acceptance: root cause identified, and either the testset restored to
HiGHS_optimizer or the comment replaced with a specific explanation of
why Ipopt is required.
Introduced in #225.