Wrong bound types when using MOSEK

[goghino]

I am trying to solve the UC using MOSEK solver, e.g.

md = ModelData.read(os.path.join(this_module_path, '..', '..', 'download',
                                 'pglib-uc-master', 'rts_gmlc', '2020-01-27.json'),
                                file_type = 'pglib-uc')

md_sol = solve_unit_commitment(md, 'mosek', mipgap=0.01, timelimit=300, solver_options={}, solver_tee=True, uc_model_generator=create_tight_unit_commitment_model)

and I get en error mosek.Error: rescode.err_fixed_bound_values(1420): For a fixed constraint/variable the specified lower bound 0.000000000000000e+00 and upper bound 1.000000000000000e+00 are different.

The problem is reproducible for other test examples and UC formulations. Any suggestions/pointers how to address this issue? Thanks!

[goghino]

Update

I have changed Pyomo code generating the bound types for MOSEK. There is an indicator in the variables array (var_seq) .is_fixed() which I ignore for the moment. I have changed file /Users/Juraj/anaconda3/envs/UnitCommitmentNew/lib/python3.8/site-packages/pyomo/solvers/plugins/solvers/mosek_direct.py , line 299 in _add_vars from

299:        fxs = tuple(v.is_fixed() for v in var_seq)

to

299:        fxs = tuple(False for v in var_seq) #tuple(v.is_fixed() for v in var_seq)

The fixed variables in the RTS_GMLS model are:

matches = (x for x in var_seq if x.is_fixed())
m = [x for x in matches]
names = [n.local_name for n in m]
names
['UnitOn[121_NUCLEAR_1_T,1]', 'UnitOn[121_NUCLEAR_1_T,2]', 'UnitOn[121_NUCLEAR_1_T,3]', 'UnitOn[121_NUCLEAR_1_T,4]', 'UnitOn[121_NUCLEAR_1_T,5]', 'UnitOn[121_NUCLEAR_1_T,6]', 'UnitOn[121_NUCLEAR_1_T,7]', 'UnitOn[121_NUCLEAR_1_T,8]', 'UnitOn[121_NUCLEAR_1_T,9]', 'UnitOn[121_NUCLEAR_1_T,10]', 'UnitOn[121_NUCLEAR_1_T,11]', 'UnitOn[121_NUCLEAR_1_T,12]', 'UnitOn[121_NUCLEAR_1_T,13]', 'UnitOn[121_NUCLEAR_1_T,14]', ...]
special variables
function variables
00:'UnitOn[121_NUCLEAR_1_T,1]'
01:'UnitOn[121_NUCLEAR_1_T,2]'
02:'UnitOn[121_NUCLEAR_1_T,3]'
03:'UnitOn[121_NUCLEAR_1_T,4]'
04:'UnitOn[121_NUCLEAR_1_T,5]'
05:'UnitOn[121_NUCLEAR_1_T,6]'
06:'UnitOn[121_NUCLEAR_1_T,7]'
07:'UnitOn[121_NUCLEAR_1_T,8]'
08:'UnitOn[121_NUCLEAR_1_T,9]'
09:'UnitOn[121_NUCLEAR_1_T,10]'
10:'UnitOn[121_NUCLEAR_1_T,11]'
11:'UnitOn[121_NUCLEAR_1_T,12]'
12:'UnitOn[121_NUCLEAR_1_T,13]'
13:'UnitOn[121_NUCLEAR_1_T,14]'
14:'UnitOn[121_NUCLEAR_1_T,15]'
15:'UnitOn[121_NUCLEAR_1_T,16]'
16:'UnitOn[121_NUCLEAR_1_T,17]'
17:'UnitOn[121_NUCLEAR_1_T,18]'
18:'UnitOn[121_NUCLEAR_1_T,19]'
19:'UnitOn[121_NUCLEAR_1_T,20]'
20:'UnitOn[121_NUCLEAR_1_T,21]'
21:'UnitOn[121_NUCLEAR_1_T,22]'
22:'UnitOn[121_NUCLEAR_1_T,23]'
23:'UnitOn[121_NUCLEAR_1_T,24]'
24:'UnitOn[121_NUCLEAR_1_T,25]'
25:'UnitOn[121_NUCLEAR_1_T,26]'
26:'UnitOn[121_NUCLEAR_1_T,27]'
27:'UnitOn[121_NUCLEAR_1_T,28]'
28:'UnitOn[121_NUCLEAR_1_T,29]'
29:'UnitOn[121_NUCLEAR_1_T,30]'
30:'UnitOn[121_NUCLEAR_1_T,31]'
31:'UnitOn[121_NUCLEAR_1_T,32]'
32:'UnitOn[121_NUCLEAR_1_T,33]'
33:'UnitOn[121_NUCLEAR_1_T,34]'
34:'UnitOn[121_NUCLEAR_1_T,35]'
35:'UnitOn[121_NUCLEAR_1_T,36]'
36:'UnitOn[121_NUCLEAR_1_T,37]'
37:'UnitOn[121_NUCLEAR_1_T,38]'
38:'UnitOn[121_NUCLEAR_1_T,39]'
39:'UnitOn[121_NUCLEAR_1_T,40]'
40:'UnitOn[121_NUCLEAR_1_T,41]'
41:'UnitOn[121_NUCLEAR_1_T,42]'
42:'UnitOn[121_NUCLEAR_1_T,43]'
43:'UnitOn[121_NUCLEAR_1_T,44]'
44:'UnitOn[121_NUCLEAR_1_T,45]'
45:'UnitOn[121_NUCLEAR_1_T,46]'
46:'UnitOn[121_NUCLEAR_1_T,47]'
47:'UnitOn[121_NUCLEAR_1_T,48]'
len():48

Then, running MOSEK works fine

$ python pglib_uc_example_juraj.py
Creating and solving pglib-uc example ...
Calculating PTDF Matrix Factorization
Problem
  Name                   :
  Objective sense        : min
  Type                   : LO (linear optimization problem)
  Constraints            : 51903
  Cones                  : 0
  Scalar variables       : 53537
  Matrix variables       : 0
  Integer variables      : 21425

Optimizer started.
Mixed integer optimizer started.
Threads used: 8
Presolve started.
Presolve terminated. Time = 1.43
Presolved problem: 39320 variables, 38393 constraints, 184453 non-zeros
Presolved problem: 1647 general integer, 21081 binary, 16592 continuous
Clique table size: 26680
BRANCHES RELAXS   ACT_NDS  DEPTH    BEST_INT_OBJ         BEST_RELAX_OBJ       REL_GAP(%)  TIME
0        1        1        0        NA                   1.2266453400e+06     NA          14.3
0        1        1        0        1.2328787585e+06     1.2266453400e+06     0.51        20.1
Cut generation started.
0        1        1        0        1.2328787585e+06     1.2266453400e+06     0.51        22.3
Cut generation terminated. Time = 0.35
Presolve started.
Presolve terminated. Time = 1.53
Presolved problem: 30970 variables, 36470 constraints, 149759 non-zeros
Presolved problem: 1617 general integer, 12773 binary, 16580 continuous
Clique table size: 25028
BRANCHES RELAXS   ACT_NDS  DEPTH    BEST_INT_OBJ         BEST_RELAX_OBJ       REL_GAP(%)  TIME
0        4        1        0        1.2328787585e+06     1.2268057436e+06     0.49        24.7
0        7        1        0        1.2328787585e+06     1.2268057436e+06     0.49        33.4
0        18       1        0        1.2328787585e+06     1.2268057436e+06     0.49        34.1
12       31       1        0        1.2328787585e+06     1.2268057436e+06     0.49        35.2
26       45       1        0        1.2328787585e+06     1.2268057436e+06     0.49        35.5
50       69       1        0        1.2328787585e+06     1.2268057436e+06     0.49        36.0

[michaelbynum]

I believe this is now fixed in Pyomo (pyomo/pyomo#1844).

[goghino]

I am using the latest Pyomo via Anaconda 3 and I still experience the issue pyomo 5.7.3 py38h91a8764_0 conda-forge

[michaelbynum]

Sorry - it was fixed in the main branch. There has not been a release since pyomo/pyomo#1844 was merged. It will probably be another month before there is another Pyomo release.