2
$\begingroup$

I'm using the Java API of CPLEX (12.6.1 version) to solve a MILP problem.

This is how I create 'normal' constraints:

public void charge_discharge_constraints() throws IloException {

    for (int k = 0; k < periods; k++) {
        for (int p = 0; p < num_p; p++) {
            IloLinearNumExpr exp = model.linearNumExpr();
            for (int n = 0; n < num_n; n++)
                exp.addTerm(1, bc_p[n][p][k]);
            exp.addTerm(1.0, bd_p[p][k]);
            IloRange constr = model.addGe(1, exp, "Non simultaneous charge discharge");
        }
    }
}

Now, I'm trying to add an 'indicator constraint' such as:

if bd_p[p][k] == 1, -> (then) h_p[p][k] >= 4800.0

for every p and k indices.

I tried to achieve this by doing:

public void indicator_constraints() throws IloException {

    for (int p = 0; p < num_p; p++) {
        for (int k = 0; k < periods; k++) {
            IloLinearNumExpr exp = model.linearNumExpr();
            exp.addTerm(1, h_p[p][k]);
            model.add(model.ifThen(model.eq(bd_p[p][k], 1.0), model.ge(exp, 4800.0)));
        }
    }
}

But, after having observed the generated ".lp" file, I noticed that the indicator constraints I created use the double-arrow simbol (<->), while my aim is to use the right-arrow symbol (->).

How could I achieve this?


EDIT (to add more details):

I can confirm that I have previously assigned a name to variables bd_p and h_p. Indeed, inside the LP file:

  • both of them are present in the Binaries section (the section which lists the binary variables of the problem);
  • they're correctly bounded between 0 and 1.

Anyway, while I was expecting that the indicator constraints which I generated should have used variables bd_p and h_p, actually (after a further investigation of the LP file) I can say that only the h_p variable is involved in the indicator constraints.

Indeed, the LP file contains indicator constraints formulated as:

IloI785: x38223#36652 = 1 <-> h_p(27,17)#34158 >= 4800

So, while h_p variable compares, bd_p variable is not present. In place of bd_p there's another variable called x followed by a progressive number.

bd_p is a binary variable, I previously assigned a name to it, it is correctly defined/bounded in the LP file, and it is used in many other "normal" (linear) constraints without any problem.

I also tried to do this small change:

public void indicator_constraints() throws IloException {  
    for (int p = 0; p < num_p; p++) {
            for (int k = 0; k < periods; k++) {
                IloLinearNumExpr exp_if = model.linearNumExpr();
                IloLinearNumExpr exp_then = model.linearNumExpr();
                exp_if.addTerm(1, bd_p[p][k]);
                exp_then.addTerm(1, h_p[p][k]);
                model.add(model.ifThen(model.eq(exp_if, 1.0), model.ge(exp_then, 4800.0)));
            }
        }
    }

or, this other alternative:

public void indicator_constraints() throws IloException {  
    for (int p = 0; p < num_p; p++) {
            for (int k = 0; k < periods; k++) {
                IloLinearNumExpr exp_if = model.linearNumExpr();
                IloLinearNumExpr exp_then = model.linearNumExpr();
                exp_if.addTerm(1, bd_p[p][k]);
                exp_then.addTerm(1, h_p[p][k]);
                IloRange constr_if = model.addEq(1.0, exp_if);
                IloRange constr_then = model.addLe(4800.0, exp_then);
                model.add(model.ifThen(constr_if, constr_then));
            }
        }
    }

but with no improvement.

So, I can't understand why in the LP file, the binary variable h_p is replaced by a unnamed variable.

Which could be the problem?

$\endgroup$
2
$\begingroup$

I suspect you have not assigned names to your variables. Try doing that and then exporting the .lp file, and I think you will discover that the bidirectional constraint involves a variable that CPLEX added.

As a small example, I created a model with a single binary variable "x", a single continuous variable "y" (domain [0, 10]), and a single constraint that x = 1 -> y <= 5. CPLEX added binary variables x1 and x3 and converted the one constraint I had to two constraints: x3 = 1 <-> y =< 5 and x = 1 -> x3 = 1. I'm not sure why it does that, but I presume the programmers had their reasons.

$\endgroup$
2
  • $\begingroup$ Thank you for your suggestions, very interesting! Actually, I have correctly previoulsy assigned a name to all my variables, but you gave me the intuition to notice that, while the h_p variable is present in the indicator constraints in the LP file, the binary variable bd_p doesn't compare (in place of bd_p there's a generic variable called x followed by a progressive number). Its strange to me; anyway, I edited my original post to add some details which can be useful to you in order to help me to solve this problem. $\endgroup$
    – rainbow
    Jun 8 at 22:09
  • $\begingroup$ IloI785 is as I would expect. There should also be another constraint somewhere saying bd_p(27,17)#<whatever> = 1 -> x38223#36652 = 1. Are you saying that it is not present anywhere in the .lp file? $\endgroup$
    – prubin
    Jun 9 at 23:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.