Why do I get Infeasible Solution (JuMP with Optimizer Ipopt)

Specific Domains Optimization (Mathematical)
21

Correct. I didn’t read clearly enough.

No, both would. @constraint(model, 0.9 <= x <= 1) adds a ScalarAffineFunction-in-Interval constraint.

22

But I imagine the order is also important: fixing before adding the constraints will result in an error/infeasibility?

23

Oops. I didn’t read your question enough. Both result in infeasibilities.

1 Like
24

So fix only applies to the bounds declared with the variable

1 Like
25

Yes. You can set variable bounds one of two ways:

@variable(model, l <= x <= u)

or

@variable(model, x)
set_lower_bound(x, l)
set_upper_bound(x, u)
1 Like
26

I am watching your educational video
Thank you
and I have another question please
If I want to validate my ACOPF results, can I use Power World Simulator for example to do so
Is it a good choice ?

27

I have been watching lecture now on convex relaxation and it explained to me the main concept so now how can I do convex relaxation for my problem…Do I have to write the code of SDP for example or can I use certain solver instead

28

I would start with one of the simplest convex relaxations. That is the SOC relaxation in the W space. The idea is, defined new voltage product variables W^R_{i,j} + i*W^I_{i,j}, which will represent the complex voltage products V_i * V_j. You go through your model and make these replacements,

|V_i||V_i| \rightarrow W_{ii} \\ |V_i||V_j| cos(\theta_i - \theta_j) \rightarrow W^R_{ij} \\ |V_i||V_j| sin(\theta_i - \theta_j) \rightarrow W^I_{ij}

You then add this constraint,

(W^R_{ij})^2 + (W^I_{ij})^2 \leq W^R_{ii}*W^I_{ii}

As a refrence implementation here is how AC and SOC models are constructed in PowerModels,

1 Like
29

I will try that…Thank you Dr. @ccoffrin

30

I will try that…Thank you Dr. @ccoffrin

31

Aha ,so I want to use the AC OPF then switch to the rectangular or maybe I will do both in parallel…so
I used Couenne with ACOPF formulations through AmplNLWriter and I got “INFEASIBLE” :frowning:
Although with the same model and equations with Ipopt, a local solution was found
could it be something with the couenne or my model ?!

32

That sounds like a tolerance issue. Under the hood, Couenne is using Ipopt to solve the continuous problems.