IIS: IIS failed due numerical instability, got status DUAL_INFEASIBLE

[datejada]

The attached LP file comes from a tiny case, we know why it is infeasible due to an error in the input data. The model analyser throws an error, but I can get the IIS from Gurobi. This is the code:

tiny-infeasible-model.zip

using JuMP
using ModelAnalyzer
using HiGHS

# Create a JuMP model from lp file from OBZ case study
model = read_from_file("tiny-infeasible-model.lp")
set_optimizer(model, HiGHS.Optimizer)
optimize!(model)

# Infeasibility analysis (if the model was infeasible)
data = ModelAnalyzer.analyze(
    ModelAnalyzer.Infeasibility.Analyzer(),
    model,
    optimizer=HiGHS.Optimizer,
)

# print report to the screen
ModelAnalyzer.summarize(data)

This is the error:

ERROR: IIS failed due numerical instability, got status DUAL_INFEASIBLE
Stacktrace:
 [1] error(s::String)
   @ Base .\error.jl:35
 [2] iis_elastic_filter(original_model::MathOptInterface.Utilities.CachingOptimizer{…}, optimizer::Type)
   @ ModelAnalyzer.Infeasibility C:\Users\tejadaarangoda\.julia\packages\ModelAnalyzer\Sta4E\src\infeasibility.jl:521
 [3] analyze(::ModelAnalyzer.Infeasibility.Analyzer, model::MathOptInterface.Utilities.CachingOptimizer{…}; optimizer::Type)
   @ ModelAnalyzer.Infeasibility C:\Users\tejadaarangoda\.julia\packages\ModelAnalyzer\Sta4E\src\infeasibility.jl:388
 [4] analyze(analyzer::ModelAnalyzer.Infeasibility.Analyzer, model::Model; kwargs::@Kwargs{optimizer::DataType})
   @ ModelAnalyzerJuMPExt C:\Users\tejadaarangoda\.julia\packages\ModelAnalyzer\Sta4E\ext\ModelAnalyzerJuMPExt\ModelAnalyzerJuMPExt.jl:18
 [5] top-level scope
   @ c:\Users\tejadaarangoda\OneDrive - TNO\00_Sandbox\TestModelAnalyzer\test-infesiable-model.jl:12
Some type information was truncated. Use `show(err)` to see complete types.

This is code when using Gurobi:

using Gurobi
set_optimizer(model, Gurobi.Optimizer)
optimize!(model)
compute_conflict!(model)
if get_attribute(model, MOI.ConflictStatus()) == MOI.CONFLICT_FOUND
    iis_model, reference_map = copy_conflict(model)
    print(iis_model)
end

This is the output:

CPU model: 12th Gen Intel(R) Core(TM) i7-1255U, instruction set [SSE2|AVX|AVX2]
Thread count: 10 physical cores, 12 logical processors, using up to 12 threads

Optimize a model with 936 rows, 396 columns and 1656 nonzeros
Model fingerprint: 0x02df7def
Coefficient statistics:
  Matrix range     [1e+00, 3e+02]
  Objective range  [4e+00, 9e+00]
  Bounds range     [0e+00, 0e+00]
  RHS range        [8e+02, 2e+03]
Presolve removed 708 rows and 276 columns
Presolve time: 0.00s

Solved in 0 iterations and 0.00 seconds (0.00 work units)
Infeasible or unbounded model

User-callback calls 36, time in user-callback 0.00 sec
Gurobi Optimizer version 12.0.2 build v12.0.2rc0 (win64 - Windows 11.0 (26100.2))

CPU model: 12th Gen Intel(R) Core(TM) i7-1255U, instruction set [SSE2|AVX|AVX2]
Thread count: 10 physical cores, 12 logical processors, using up to 12 threads

Iteration    Objective       Primal Inf.    Dual Inf.      Time
       0      handle free variables                          0s

IIS computed: 6 constraints and 0 bounds
IIS runtime: 0.00 seconds (0.00 work units)
Min 3.6500000000000004 flow_(ccgt,demand),2030,1,1_1_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,2_2_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,3_3_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,4_4_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,5_5_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,6_6_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,7_7_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,8_8_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,9_9_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,10_10_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,11_11_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,12_12_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,13_13_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,14_14_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,15_15_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,16_16_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,17_17_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,18_18_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,19_19_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,20_20_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,21_21_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,22_22_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,23_23_ + 3.6500000000000004 flow_(ccgt,demand),2030,1,24_24_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,1_1_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,2_2_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,3_3_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,4_4_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,5_5_ + 3.6500000000000004 flow_(ccgt,hub),2030,1,6_6_ + [[...84 terms omitted...]] + 5.4750000000000005 flow_(ccgt,demand),2030,3,19_19_ + 5.4750000000000005 flow_(ccgt,demand),2030,3,20_20_ + 5.4750000000000005 flow_(ccgt,demand),2030,3,21_21_ + 5.4750000000000005 flow_(ccgt,demand),2030,3,22_22_ + 5.4750000000000005 flow_(ccgt,demand),2030,3,23_23_ + 5.4750000000000005 flow_(ccgt,demand),2030,3,24_24_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,1_1_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,2_2_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,3_3_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,4_4_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,5_5_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,6_6_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,7_7_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,8_8_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,9_9_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,10_10_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,11_11_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,12_12_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,13_13_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,14_14_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,15_15_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,16_16_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,17_17_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,18_18_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,19_19_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,20_20_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,21_21_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,22_22_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,23_23_ + 5.4750000000000005 flow_(ccgt,hub),2030,3,24_24_
Subject to
 consumer_balance_demand,2030,1,1_1_ : flow_(ccgt,demand),2030,1,1_1_ + flow_(hub,demand),2030,1,1_1_ == 949.999999755
 consumer_balance_demand,2030,1,2_2_ : flow_(ccgt,demand),2030,1,2_2_ + flow_(hub,demand),2030,1,2_2_ == 869.99999967
 dc_power_flow_ccgt,demand,2030,1,1_1_ : flow_(ccgt,demand),2030,1,1_1_ + 333.33333333333337 _electricity_angle_demand,2030,1,1_2_ - 333.33333333333337 _electricity_angle_ccgt,2030,1,1_2_ == 0
 dc_power_flow_ccgt,demand,2030,1,2_2_ : flow_(ccgt,demand),2030,1,2_2_ + 333.33333333333337 _electricity_angle_demand,2030,1,1_2_ - 333.33333333333337 _electricity_angle_ccgt,2030,1,1_2_ == 0
 dc_power_flow_hub,demand,2030,1,1_1_ : flow_(hub,demand),2030,1,1_1_ + 333.33333333333337 _electricity_angle_demand,2030,1,1_2_ - 333.33333333333337 _electricity_angle_hub,2030,1,1_2_ == 0
 dc_power_flow_hub,demand,2030,1,2_2_ : flow_(hub,demand),2030,1,2_2_ + 333.33333333333337 _electricity_angle_demand,2030,1,1_2_ - 333.33333333333337 _electricity_angle_hub,2030,1,1_2_ == 0

[joaquimg]

This was a good catch, the relaxed model can be unbounded, so we needed to treat this case. The IIS found is different, but seems to be the same underlying issue for a alternate time step.