There are some doubts about the results of the experiments in the Powermodels documentation

Specific Domains Optimization (Mathematical)
1

I would like to consult the following documents about the experimental results, for example, the calculation time of the SoC model of case1354_pegase is 3 seconds, but I calculate the time on my personal computer is 17 seconds, and I am not sure why there is such a big gap. Here is some of the code I used
`using JuMP, PowerModels, Ipopt
file_name = “E:/Vscode——data/data/case1354pegase.m”
model = Model(Ipopt.Optimizer)
power_model = PowerModels.instantiate_model(
PowerModels.parse_file(file_name),
PowerModels.SOCWRPowerModel,
PowerModels.build_opf;
jump_model = model,
)
optimize!(model)
println("Optimal objective value: ", JuMP.objective_value(model))
println("Optimal solve time: ", JuMP.solve_time(model))

`

2

The experiment that is put in the PGLib table is run like the following,

result = solve_opf("data_file.m", SOCWRPowerModel, Ipopt.Optimizer)
result["solve_time"]

The key details for timing are,

  1. make sure that the run does not include Julia’s JIT compile time. Running the function twice is sufficient
  2. For large cases (say >1000 buses) the linear solver can make a big difference in Ipopt’s performance. In these experiment HSL ma27 is used.
1 Like