Hi all,

I am working on upgrading the CPLEX version from 12.9.x to 22.1.2.
After making the adjustments and starting the tests of the new version of CPLEX (22), for comparison purposes I ran the same process (using the same data) on both versions of CPLEX and encountered some errors in the new, more recent version, errors that do not occur in the old version.

Could someone help me to understand?

Piece of logs from CPLEX 22.1.2:

TRACE : [Improving solution] Performing Solve
TRACE : [Improving solution] SetCplexParameter(EpGap, 0.0001)
TRACE : [Improving solution] SetCplexParameter(NodeLim, 1000)
TRACE : [Improving solution] SetCplexParameter(IntSolLim, 1000)
TRACE : [Improving solution] Time to prepare problem 7.5671694 seconds
TRACE : [Improving solution] SetCplexParameter(Cplex.Param.TimeLimit, 600)
TRACE : [Improving solution] SetCplexParameter(Cplex.Param.TimeLimit, 999999)
TRACE : [Improving solution] CPLEX time limit is  999999
TRACE : [Improving solution] CPLEX node limit is  1000
TRACE : [Improving solution] CPLEX soln limit is  1000
TRACE : [Improving solution] CPLEX gap limit is   0.0001
TRACE : [Improving solution] CPLEX polish time is 1E+75
TRACE : [Improving solution] CPLEX threads limit  1
Version identifier: 22.1.2.0 | 2024-11-25 | 0edbb82fd
CPXPARAM_LPMethod                                2
CPXPARAM_QPMethod                                2
CPXPARAM_Threads                                 1
CPXPARAM_Parallel                                1
CPXPARAM_MIP_Display                             4
CPXPARAM_MIP_Strategy_StartAlgorithm             2
CPXPARAM_Emphasis_MIP                            1
CPXPARAM_MIP_Strategy_Search                     1
CPXPARAM_MIP_Interval                            10
CPXPARAM_MIP_Limits_Solutions                    1000
CPXPARAM_MIP_Limits_Nodes                        1000
CPXPARAM_MIP_Strategy_HeuristicFreq              100
CPXPARAM_TimeLimit                               999999
CPXPARAM_MIP_Tolerances_AbsMIPGap                0.0001
Infeasibility row 'CtPeakDPHardLimit_C-3727484_BG_-3727484/16295976_84289':  0  = 1.
Presolve time = 0.23 sec. (201.86 ticks)

Root node processing (before b&c):
  Real time             =    0.31 sec. (253.58 ticks)
Sequential b&c:
  Real time             =    0.00 sec. (0.00 ticks)
                          ------------
Total (root+branch&cut) =    0.31 sec. (253.58 ticks)
ERROR : [Improving solution] Cplex.Solve() failed: 2 infeasible steps and 2017 feasible steps so far, File OptimisationModel.cs, Method Solve, Line 10849, Stage Solve, Description Cplex solve failed
Warning:  Output names have been modified due to duplicate names or characters invalid in LP format.
TRACE : [Improving solution] Local run completed after 9.0393297 seconds, status = False
TRACE : [Improving solution] Elapsed time after local run completed = 07:15:40.5808144
TRACE : [Improving solution] New solution found after 9.0393297    seconds
TRACE : [Improving solution] Elapsed time After Solve = 07:15:40.5808144
TRACE : [Improving solution] Impossible to solve sub problem 
TRACE : [Improving solution] ==============================================================
TRACE : [Improving solution] Writing run status ABORTED 

Piece of logs from CPLEX 12.9.x:

From the log, you get a constraint that is infeasible: Infeasibility row 'CtPeakDPHardLimit_C-3727484_BG_-3727484/16295976_84289':  0  = 1. 

If you want to have more context, you can use the tool "conflict" in CPLEX Interactive. 

Such a difference in behavior between two versions of CPLEX suggests that the model is numerically instable.