Deal with quadratic constraints

Ongoing

Lorenz Wagner

• February 11, 2024 12:36

Hello, I have the following problem. I am setting up a linear programme (or so I thought), but my model output shows something like this.

Model has xxxx quadratic constraints

Is there any way I can find out which of my constraints is causing these problems?

I have a hunch that it is this constraint here. Is there possibly a way to formulate it as a linear constraint?

for i in I:
    for t in range(1,len(T)):
        m.addConstr(n[i, t] == (n[i, t - 1] + sc[i, t]) * (1 - r[i, t]) - e[i, t] + b[i, t])

\(n_{it}\) is an integer and both \(sc_{it}\) and \(r_{it}\) are binary.

 

Thank you for your help!

Related to

• QCP

0

• 

  Riley Clement

  Gurobi Staff

  • February 14, 2024 12:52

  Hi Lorenz,

  You can retrieve quadratic constraints with gurobipy using the Model.getQConstrs() method.  You can then query relevant attributes listed here: Quadratic Constraint Attributes

  You can control the linearization of quadratic constraints (by Gurobi) using PreQLinearize.  Is there a motivation for wanting to do the linearization yourself?

  - Riley

  0
• 

  Lorenz Wagner

  • February 14, 2024 15:31

  

  

  

  Hi Riley Clement thanks a lot. My guess would be something like this.

  for i in I:
      for t in range(2, len(T) + 1):
          m.addConstr(n[i, t] == n_h[i, t] - e[i, t] + b[i, t])
          m.addConstr(n_h[i, t] <= n[i, t-1] + sc[i, t])
          m.addConstr(n_h[i, t] >= (n[i, t-1] + sc[i, t])-M*r[i, t])
          m.addConstr(n_h[i, t] <= M*(1 - r[i, t]))

  Sadly i cant test it right now. Would that work?

  0

Please sign in to leave a comment.