Too much time when i solve MISOCP with Gurobi

Specific Domains Optimization (Mathematical)
1

this is my model

julia> model
A JuMP Model
Minimization problem with:
Variables: 8915
Objective function type: AffExpr
`AffExpr`-in-`MathOptInterface.EqualTo{Float64}`: 2716 constraints
`AffExpr`-in-`MathOptInterface.GreaterThan{Float64}`: 813 constraints
`AffExpr`-in-`MathOptInterface.LessThan{Float64}`: 576 constraints
`Vector{AffExpr}`-in-`MathOptInterface.SecondOrderCone`: 768 constraints
`VariableRef`-in-`MathOptInterface.EqualTo{Float64}`: 2619 constraints
`VariableRef`-in-`MathOptInterface.GreaterThan{Float64}`: 3698 constraints
`VariableRef`-in-`MathOptInterface.LessThan{Float64}`: 3602 constraints
`VariableRef`-in-`MathOptInterface.ZeroOne`: 801 constraints
Model mode: AUTOMATIC
CachingOptimizer state: EMPTY_OPTIMIZER
Solver name: Gurobi
Names registered in the model: E_ess, I, P, P_ch, P_dch, Pg, Q, Qg, SOCP, V, c10, c16, c17_1, c17_2, c17_3, c17_4, c17_5, c18, c18_3, c18_4, c18_5, c18_6, c22, c23_1, c23_2, c24_1, c24_3, ch, ch2, dch, dch2, p_ch, p_dch, p_wt, powerbalanced, q_SVC, qpowerbalanced, theta1_DE, theta1_IN, theta_CB, theta_DE, theta_IN, theta_OLTC, u, u1, u2, u_ch, u_dch, z, z1, z2, z3, z4, zx, zx1, zx2, zx3, zx4

but it takes exponential time , how can i make it faster

    Nodes    |    Current Node    |     Objective Bounds      |     Work
 Expl Unexpl |  Obj  Depth IntInf | Incumbent    BestBd   Gap | It/Node Time

     0     0   13.54863    0  141          -   13.54863      -     -    0s
     0     0   15.86604    0  254          -   15.86604      -     -    0s
     0     0   15.95797    0  301          -   15.95797      -     -    0s
     0     0   15.97053    0  305          -   15.97053      -     -    0s
     0     0   16.01988    0  292          -   16.01988      -     -    1s
     0     0   16.02265    0  296          -   16.02265      -     -    1s
     0     0   16.04140    0  266          -   16.04140      -     -    1s
     0     0   16.04209    0  269          -   16.04209      -     -    1s
     0     0   16.05020    0  244          -   16.05020      -     -    1s
H    0     0                      18.1401667   16.05246  11.5%     -    1s
     0     0   16.05246    0  243   18.14017   16.05246  11.5%     -    1s
     0     0   16.05420    0  259   18.14017   16.05420  11.5%     -    2s
     0     0   16.05476    0  247   18.14017   16.05476  11.5%     -    2s
     0     0   16.05506    0  275   18.14017   16.05506  11.5%     -    2s
     0     0   16.05506    0  275   18.14017   16.05506  11.5%     -    2s
     0     2   16.05506    0  275   18.14017   16.05506  11.5%     -    3s
    84    98   17.84563   15  240   18.14017   16.18759  10.8%   201    5s
H  279   260                      18.0158998   16.18759  10.1%   107    5s
H  454   366                      18.0096778   16.24597  9.79%   106    7s
   461   369   17.95981   88  197   18.00968   16.24597  9.79%   106   10s
   469   374   16.80619    9  201   18.00968   16.40546  8.91%   104   15s
   477   380   17.95752   69  186   18.00968   16.40886  8.89%   102   20s
   484   386   16.43441   16  200   18.00968   16.43441  8.75%   151   32s
   497   397   16.69291   19  209   18.00968   16.48738  8.45%   161   35s
   571   443   17.28456   29  235   18.00968   16.52335  8.25%   174   40s
  .
  .
  .
 801954 130763   17.91647   42  132   17.94462   17.91041  0.19%   218 47687s
 802597 130806     cutoff   45        17.94462   17.91042  0.19%   218 47742s
 803147 130858   17.91256   41  132   17.94462   17.91043  0.19%   218 47785s
1 Like
2

Do you have a reproducible example?

Otherwise you can set a time limit or change the optimality gap so that Gurobi terminates earlier. MISOCPs are hard to solve, even relatively small ones like this.

3

i wan to change the optimality gap , how to code to achieve it? i didn`t find related function

1 Like
4 
model = Model(Gurobi.Optimizer)
set_optimizer_attribute(model, "MIPGap", 0.05)

Gurobi parameters are here: https://www.gurobi.com/documentation/9.5/refman/parameters.html

5

thank you very much

1 Like